Search Results

Search found 39999 results on 1600 pages for 'php session'.

Page 36/1600 | < Previous Page | 32 33 34 35 36 37 38 39 40 41 42 43  | Next Page >

  • I can not use Session In Page_Load and I got error bellow

    - by LostLord
    hi my dear friends .... why i got this error : Object reference not set to an instance of an object. when i put this code in my page_load.: protected void Page_Load(object sender, EventArgs e) { BackEndUtils.OverallLoader(); string Teststr = Session["Co_ID"].ToString(); } ========================================================================== this session is made when user logins to my web site and this session works in other areas... thanks for your attention ========================================================================== thanks for your answers i removed BackEndUtils.OverallLoader(); but error still exists i tried Teststr = Convert.ToString(Session["Co_ID"]); and error disappeared - but i don't know why that session is null in other areas that session works perfectly = such as a button in that form what is the matter? my web page markup is like this : <%@ Page Title="" Language="C#" MasterPageFile="~/Admin/AdminBackend.Master" AutoEventWireup="true" CodeBehind="Personel.aspx.cs" Inherits="Darman.Admin.Personel" Theme="DefaultTheme" %> ================================================================================= i put this code in a button like this : string Teststr = Convert.ToString(Session["Co_ID"]); when i press that button THIS code in page Load(POSTBACK) + IN Button_Click works perfectly and shows me 23 (my Co_ID) But when i run my page in browser (first time) this code in page load shows me null. why? thanks a lot

    Read the article

  • webpart context.session is null

    - by tbischel
    I've been using the session array to store a state variable for my webpart... so I have a property like this: public INode RootNode { get { return this.Context.Session["RootNode"] as INode; } set { this.Context.Session["RootNode"] = value as object; } } This usually works fine. I've discovered that sometimes, the context.session variable will be null. I'd like to know what are the conditions that cause the session to be null in the first place, and whats the best way to persist my object when this happens? Can I just assign a new HttpSessionState object to the context, or does that screw things up? Edit: Ok, so its not just the session that is null... the whole context is screwed up. When the webpart enters the init, the context is fine... but when it reaches the dropbox selectedindexchange postback event (the dropbox contains node id's to use to set the rootnode variable), the context contains mostly null properties. also, it only seems to happen when certain id's are selected. This looks more like some kind of weird bug on my end than a problem with my understanding of the session.

    Read the article

  • Mysterious HttpSession and session-config dependency

    - by OneMoreVladimir
    Good day. I'm developing a Java web app with Servlets\JSP using Tomcat 7.0. During request from client I put and object into the session and use forward. After the forward processing the same request the object can be retreived if the secure parameter is false otherwise it is not stored in session. <session-config> <session-timeout>15</session-timeout> <cookie-config> <http-only>true</http-only> <secure>true</secure> </cookie-config> <tracking-mode>COOKIE</tracking-mode> </session-config> I've figured out that "...cookies can be created with the 'secure' flag, which ensures that the browser will never transmit the specified cookie over non-SSL...". I've configured Tomcat to use SSL, but that haven't helped. Changing the tracking mode to SSL haven't helped as well. How do session-config and HttpSession object correlate in this case? What could be the problem?

    Read the article

  • Option Trading: Getting the most out of the event session options

    - by extended_events
    You can control different aspects of how an event session behaves by setting the event session options as part of the CREATE EVENT SESSION DDL. The default settings for the event session options are designed to handle most of the common event collection situations so I generally recommend that you just use the defaults. Like everything in the real world though, there are going to be a handful of “special cases” that require something different. This post focuses on identifying the special cases and the correct use of the options to accommodate those cases. There is a reason it’s called Default The default session options specify a total event buffer size of 4 MB with a 30 second latency. Translating this into human terms; this means that our default behavior is that the system will start processing events from the event buffer when we reach about 1.3 MB of events or after 30 seconds, which ever comes first. Aside: What’s up with the 1.3 MB, I thought you said the buffer was 4 MB?The Extended Events engine takes the total buffer size specified by MAX_MEMORY (4MB by default) and divides it into 3 equally sized buffers. This is done so that a session can be publishing events to one buffer while other buffers are being processed. There are always at least three buffers; how to get more than three is covered later. Using this configuration, the Extended Events engine can “keep up” with most event sessions on standard workloads. Why is this? The fact is that most events are small, really small; on the order of a couple hundred bytes. Even when you start considering events that carry dynamically sized data (eg. binary, text, etc.) or adding actions that collect additional data, the total size of the event is still likely to be pretty small. This means that each buffer can likely hold thousands of events before it has to be processed. When the event buffers are finally processed there is an economy of scale achieved since most targets support bulk processing of the events so they are processed at the buffer level rather than the individual event level. When all this is working together it’s more likely that a full buffer will be processed and put back into the ready queue before the remaining buffers (remember, there are at least three) are full. I know what you’re going to say: “My server is exceptional! My workload is so massive it defies categorization!” OK, maybe you weren’t going to say that exactly, but you were probably thinking it. The point is that there are situations that won’t be covered by the Default, but that’s a good place to start and this post assumes you’ve started there so that you have something to look at in order to determine if you do have a special case that needs different settings. So let’s get to the special cases… What event just fired?! How about now?! Now?! If you believe the commercial adage from Heinz Ketchup (Heinz Slow Good Ketchup ad on You Tube), some things are worth the wait. This is not a belief held by most DBAs, particularly DBAs who are looking for an answer to a troubleshooting question fast. If you’re one of these anxious DBAs, or maybe just a Program Manager doing a demo, then 30 seconds might be longer than you’re comfortable waiting. If you find yourself in this situation then consider changing the MAX_DISPATCH_LATENCY option for your event session. This option will force the event buffers to be processed based on your time schedule. This option only makes sense for the asynchronous targets since those are the ones where we allow events to build up in the event buffer – if you’re using one of the synchronous targets this option isn’t relevant. Avoid forgotten events by increasing your memory Have you ever had one of those days where you keep forgetting things? That can happen in Extended Events too; we call it dropped events. In order to optimizes for server performance and help ensure that the Extended Events doesn’t block the server if to drop events that can’t be published to a buffer because the buffer is full. You can determine if events are being dropped from a session by querying the dm_xe_sessions DMV and looking at the dropped_event_count field. Aside: Should you care if you’re dropping events?Maybe not – think about why you’re collecting data in the first place and whether you’re really going to miss a few dropped events. For example, if you’re collecting query duration stats over thousands of executions of a query it won’t make a huge difference to miss a couple executions. Use your best judgment. If you find that your session is dropping events it means that the event buffer is not large enough to handle the volume of events that are being published. There are two ways to address this problem. First, you could collect fewer events – examine you session to see if you are over collecting. Do you need all the actions you’ve specified? Could you apply a predicate to be more specific about when you fire the event? Assuming the session is defined correctly, the next option is to change the MAX_MEMORY option to a larger number. Picking the right event buffer size might take some trial and error, but a good place to start is with the number of dropped events compared to the number you’ve collected. Aside: There are three different behaviors for dropping events that you specify using the EVENT_RETENTION_MODE option. The default is to allow single event loss and you should stick with this setting since it is the best choice for keeping the impact on server performance low.You’ll be tempted to use the setting to not lose any events (NO_EVENT_LOSS) – resist this urge since it can result in blocking on the server. If you’re worried that you’re losing events you should be increasing your event buffer memory as described in this section. Some events are too big to fail A less common reason for dropping an event is when an event is so large that it can’t fit into the event buffer. Even though most events are going to be small, you might find a condition that occasionally generates a very large event. You can determine if your session is dropping large events by looking at the dm_xe_sessions DMV once again, this time check the largest_event_dropped_size. If this value is larger than the size of your event buffer [remember, the size of your event buffer, by default, is max_memory / 3] then you need a large event buffer. To specify a large event buffer you set the MAX_EVENT_SIZE option to a value large enough to fit the largest event dropped based on data from the DMV. When you set this option the Extended Events engine will create two buffers of this size to accommodate these large events. As an added bonus (no extra charge) the large event buffer will also be used to store normal events in the cases where the normal event buffers are all full and waiting to be processed. (Note: This is just a side-effect, not the intended use. If you’re dropping many normal events then you should increase your normal event buffer size.) Partitioning: moving your events to a sub-division Earlier I alluded to the fact that you can configure your event session to use more than the standard three event buffers – this is called partitioning and is controlled by the MEMORY_PARTITION_MODE option. The result of setting this option is fairly easy to explain, but knowing when to use it is a bit more art than science. First the science… You can configure partitioning in three ways: None, Per NUMA Node & Per CPU. This specifies the location where sets of event buffers are created with fairly obvious implication. There are rules we follow for sub-dividing the total memory (specified by MAX_MEMORY) between all the event buffers that are specific to the mode used: None: 3 buffers (fixed)Node: 3 * number_of_nodesCPU: 2.5 * number_of_cpus Here are some examples of what this means for different Node/CPU counts: Configuration None Node CPU 2 CPUs, 1 Node 3 buffers 3 buffers 5 buffers 6 CPUs, 2 Node 3 buffers 6 buffers 15 buffers 40 CPUs, 5 Nodes 3 buffers 15 buffers 100 buffers   Aside: Buffer size on multi-processor computersAs the number of Nodes or CPUs increases, the size of the event buffer gets smaller because the total memory is sub-divided into more pieces. The defaults will hold up to this for a while since each buffer set is holding events only from the Node or CPU that it is associated with, but at some point the buffers will get too small and you’ll either see events being dropped or you’ll get an error when you create your session because you’re below the minimum buffer size. Increase the MAX_MEMORY setting to an appropriate number for the configuration. The most likely reason to start partitioning is going to be related to performance. If you notice that running an event session is impacting the performance of your server beyond a reasonably expected level [Yes, there is a reasonably expected level of work required to collect events.] then partitioning might be an answer. Before you partition you might want to check a few other things: Is your event retention set to NO_EVENT_LOSS and causing blocking? (I told you not to do this.) Consider changing your event loss mode or increasing memory. Are you over collecting and causing more work than necessary? Consider adding predicates to events or removing unnecessary events and actions from your session. Are you writing the file target to the same slow disk that you use for TempDB and your other high activity databases? <kidding> <not really> It’s always worth considering the end to end picture – if you’re writing events to a file you can be impacted by I/O, network; all the usual stuff. Assuming you’ve ruled out the obvious (and not so obvious) issues, there are performance conditions that will be addressed by partitioning. For example, it’s possible to have a successful event session (eg. no dropped events) but still see a performance impact because you have many CPUs all attempting to write to the same free buffer and having to wait in line to finish their work. This is a case where partitioning would relieve the contention between the different CPUs and likely reduce the performance impact cause by the event session. There is no DMV you can check to find these conditions – sorry – that’s where the art comes in. This is  largely a matter of experimentation. On the bright side you probably won’t need to to worry about this level of detail all that often. The performance impact of Extended Events is significantly lower than what you may be used to with SQL Trace. You will likely only care about the impact if you are trying to set up a long running event session that will be part of your everyday workload – sessions used for short term troubleshooting will likely fall into the “reasonably expected impact” category. Hey buddy – I think you forgot something OK, there are two options I didn’t cover: STARTUP_STATE & TRACK_CAUSALITY. If you want your event sessions to start automatically when the server starts, set the STARTUP_STATE option to ON. (Now there is only one option I didn’t cover.) I’m going to leave causality for another post since it’s not really related to session behavior, it’s more about event analysis. - Mike Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!

    Read the article

  • Unity launcher doesn't appear in VNC session

    - by Lorin Hochstein
    I'd like to have a unity desktop accessible via VNC on a precise server machine I have running. I installed the ubuntu-desktop package. I'm launching a VNC session with the following in my ~/.vnc/xstartup file: !/bin/sh xrdb $HOME/.Xresources xsetroot -solid grey export XKL_XMODMAP_DISABLE=1 gnome-session --session=ubuntu-2d What could the issue be here? The content of my /usr/share/gnome-session/sessions/ubuntu-2d.session file (this is what 12.04 puts there by default) is: [GNOME Session] Name=Ubuntu 2D RequiredComponents=gnome-settings-daemon; RequiredProviders=windowmanager;panel;shell; DefaultProvider-windowmanager=metacity DefaultProvider-panel=unity-2d-panel DefaultProvider-shell=unity-2d-shell DesktopName=Unity I don't know anything about unity-2d, but I guessed that the default session was not configured to use the launcher. I tried making the following modifications: [GNOME Session] Name=Ubuntu 2D RequiredComponents=gnome-settings-daemon; RequiredProviders=windowmanager;panel;shell;launcher; DefaultProvider-windowmanager=metacity DefaultProvider-panel=unity-2d-panel DefaultProvider-shell=unity-2d-shell DefaultProvider-launcher=unity-2d-launcher DesktopName=Unity But I got the following warning in my VNC log file: gnome-session[10354]: WARNING: Unable to find default provider 'qr' of required provider 'launcher' And the VNC session wouldn't even start with that configuration.

    Read the article

  • Option Trading: Getting the most out of the event session options

    - by extended_events
    You can control different aspects of how an event session behaves by setting the event session options as part of the CREATE EVENT SESSION DDL. The default settings for the event session options are designed to handle most of the common event collection situations so I generally recommend that you just use the defaults. Like everything in the real world though, there are going to be a handful of “special cases” that require something different. This post focuses on identifying the special cases and the correct use of the options to accommodate those cases. There is a reason it’s called Default The default session options specify a total event buffer size of 4 MB with a 30 second latency. Translating this into human terms; this means that our default behavior is that the system will start processing events from the event buffer when we reach about 1.3 MB of events or after 30 seconds, which ever comes first. Aside: What’s up with the 1.3 MB, I thought you said the buffer was 4 MB?The Extended Events engine takes the total buffer size specified by MAX_MEMORY (4MB by default) and divides it into 3 equally sized buffers. This is done so that a session can be publishing events to one buffer while other buffers are being processed. There are always at least three buffers; how to get more than three is covered later. Using this configuration, the Extended Events engine can “keep up” with most event sessions on standard workloads. Why is this? The fact is that most events are small, really small; on the order of a couple hundred bytes. Even when you start considering events that carry dynamically sized data (eg. binary, text, etc.) or adding actions that collect additional data, the total size of the event is still likely to be pretty small. This means that each buffer can likely hold thousands of events before it has to be processed. When the event buffers are finally processed there is an economy of scale achieved since most targets support bulk processing of the events so they are processed at the buffer level rather than the individual event level. When all this is working together it’s more likely that a full buffer will be processed and put back into the ready queue before the remaining buffers (remember, there are at least three) are full. I know what you’re going to say: “My server is exceptional! My workload is so massive it defies categorization!” OK, maybe you weren’t going to say that exactly, but you were probably thinking it. The point is that there are situations that won’t be covered by the Default, but that’s a good place to start and this post assumes you’ve started there so that you have something to look at in order to determine if you do have a special case that needs different settings. So let’s get to the special cases… What event just fired?! How about now?! Now?! If you believe the commercial adage from Heinz Ketchup (Heinz Slow Good Ketchup ad on You Tube), some things are worth the wait. This is not a belief held by most DBAs, particularly DBAs who are looking for an answer to a troubleshooting question fast. If you’re one of these anxious DBAs, or maybe just a Program Manager doing a demo, then 30 seconds might be longer than you’re comfortable waiting. If you find yourself in this situation then consider changing the MAX_DISPATCH_LATENCY option for your event session. This option will force the event buffers to be processed based on your time schedule. This option only makes sense for the asynchronous targets since those are the ones where we allow events to build up in the event buffer – if you’re using one of the synchronous targets this option isn’t relevant. Avoid forgotten events by increasing your memory Have you ever had one of those days where you keep forgetting things? That can happen in Extended Events too; we call it dropped events. In order to optimizes for server performance and help ensure that the Extended Events doesn’t block the server if to drop events that can’t be published to a buffer because the buffer is full. You can determine if events are being dropped from a session by querying the dm_xe_sessions DMV and looking at the dropped_event_count field. Aside: Should you care if you’re dropping events?Maybe not – think about why you’re collecting data in the first place and whether you’re really going to miss a few dropped events. For example, if you’re collecting query duration stats over thousands of executions of a query it won’t make a huge difference to miss a couple executions. Use your best judgment. If you find that your session is dropping events it means that the event buffer is not large enough to handle the volume of events that are being published. There are two ways to address this problem. First, you could collect fewer events – examine you session to see if you are over collecting. Do you need all the actions you’ve specified? Could you apply a predicate to be more specific about when you fire the event? Assuming the session is defined correctly, the next option is to change the MAX_MEMORY option to a larger number. Picking the right event buffer size might take some trial and error, but a good place to start is with the number of dropped events compared to the number you’ve collected. Aside: There are three different behaviors for dropping events that you specify using the EVENT_RETENTION_MODE option. The default is to allow single event loss and you should stick with this setting since it is the best choice for keeping the impact on server performance low.You’ll be tempted to use the setting to not lose any events (NO_EVENT_LOSS) – resist this urge since it can result in blocking on the server. If you’re worried that you’re losing events you should be increasing your event buffer memory as described in this section. Some events are too big to fail A less common reason for dropping an event is when an event is so large that it can’t fit into the event buffer. Even though most events are going to be small, you might find a condition that occasionally generates a very large event. You can determine if your session is dropping large events by looking at the dm_xe_sessions DMV once again, this time check the largest_event_dropped_size. If this value is larger than the size of your event buffer [remember, the size of your event buffer, by default, is max_memory / 3] then you need a large event buffer. To specify a large event buffer you set the MAX_EVENT_SIZE option to a value large enough to fit the largest event dropped based on data from the DMV. When you set this option the Extended Events engine will create two buffers of this size to accommodate these large events. As an added bonus (no extra charge) the large event buffer will also be used to store normal events in the cases where the normal event buffers are all full and waiting to be processed. (Note: This is just a side-effect, not the intended use. If you’re dropping many normal events then you should increase your normal event buffer size.) Partitioning: moving your events to a sub-division Earlier I alluded to the fact that you can configure your event session to use more than the standard three event buffers – this is called partitioning and is controlled by the MEMORY_PARTITION_MODE option. The result of setting this option is fairly easy to explain, but knowing when to use it is a bit more art than science. First the science… You can configure partitioning in three ways: None, Per NUMA Node & Per CPU. This specifies the location where sets of event buffers are created with fairly obvious implication. There are rules we follow for sub-dividing the total memory (specified by MAX_MEMORY) between all the event buffers that are specific to the mode used: None: 3 buffers (fixed)Node: 3 * number_of_nodesCPU: 2.5 * number_of_cpus Here are some examples of what this means for different Node/CPU counts: Configuration None Node CPU 2 CPUs, 1 Node 3 buffers 3 buffers 5 buffers 6 CPUs, 2 Node 3 buffers 6 buffers 15 buffers 40 CPUs, 5 Nodes 3 buffers 15 buffers 100 buffers   Aside: Buffer size on multi-processor computersAs the number of Nodes or CPUs increases, the size of the event buffer gets smaller because the total memory is sub-divided into more pieces. The defaults will hold up to this for a while since each buffer set is holding events only from the Node or CPU that it is associated with, but at some point the buffers will get too small and you’ll either see events being dropped or you’ll get an error when you create your session because you’re below the minimum buffer size. Increase the MAX_MEMORY setting to an appropriate number for the configuration. The most likely reason to start partitioning is going to be related to performance. If you notice that running an event session is impacting the performance of your server beyond a reasonably expected level [Yes, there is a reasonably expected level of work required to collect events.] then partitioning might be an answer. Before you partition you might want to check a few other things: Is your event retention set to NO_EVENT_LOSS and causing blocking? (I told you not to do this.) Consider changing your event loss mode or increasing memory. Are you over collecting and causing more work than necessary? Consider adding predicates to events or removing unnecessary events and actions from your session. Are you writing the file target to the same slow disk that you use for TempDB and your other high activity databases? <kidding> <not really> It’s always worth considering the end to end picture – if you’re writing events to a file you can be impacted by I/O, network; all the usual stuff. Assuming you’ve ruled out the obvious (and not so obvious) issues, there are performance conditions that will be addressed by partitioning. For example, it’s possible to have a successful event session (eg. no dropped events) but still see a performance impact because you have many CPUs all attempting to write to the same free buffer and having to wait in line to finish their work. This is a case where partitioning would relieve the contention between the different CPUs and likely reduce the performance impact cause by the event session. There is no DMV you can check to find these conditions – sorry – that’s where the art comes in. This is  largely a matter of experimentation. On the bright side you probably won’t need to to worry about this level of detail all that often. The performance impact of Extended Events is significantly lower than what you may be used to with SQL Trace. You will likely only care about the impact if you are trying to set up a long running event session that will be part of your everyday workload – sessions used for short term troubleshooting will likely fall into the “reasonably expected impact” category. Hey buddy – I think you forgot something OK, there are two options I didn’t cover: STARTUP_STATE & TRACK_CAUSALITY. If you want your event sessions to start automatically when the server starts, set the STARTUP_STATE option to ON. (Now there is only one option I didn’t cover.) I’m going to leave causality for another post since it’s not really related to session behavior, it’s more about event analysis. - Mike Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!

    Read the article

  • Converting Lighttpd config to NginX with php-fpm

    - by Le Dude
    Having so much issue with NginX configuration since I'm new with NginX. Been using Lighttpd for quite sometime. Here are the base info. New Machine - CentOS 6.3 64 Bit - NginX 1.2.4-1.e16.ngx - Php-FPM 5.3.18-1.e16.remi Old Machine - CentOS 6.2 64Bit - Lighttpd 1.4.25-3.e16 Original Lighttpd config file: ####################################################################### ## ## /etc/lighttpd/lighttpd.conf ## ## check /etc/lighttpd/conf.d/*.conf for the configuration of modules. ## ####################################################################### ####################################################################### ## ## Some Variable definition which will make chrooting easier. ## ## if you add a variable here. Add the corresponding variable in the ## chroot example aswell. ## var.log_root = "/var/log/lighttpd" var.server_root = "/var/www" var.state_dir = "/var/run" var.home_dir = "/var/lib/lighttpd" var.conf_dir = "/etc/lighttpd" ## ## run the server chrooted. ## ## This requires root permissions during startup. ## ## If you run Chrooted set the the variables to directories relative to ## the chroot dir. ## ## example chroot configuration: ## #var.log_root = "/logs" #var.server_root = "/" #var.state_dir = "/run" #var.home_dir = "/lib/lighttpd" #var.vhosts_dir = "/vhosts" #var.conf_dir = "/etc" # #server.chroot = "/srv/www" ## ## Some additional variables to make the configuration easier ## ## ## Base directory for all virtual hosts ## ## used in: ## conf.d/evhost.conf ## conf.d/simple_vhost.conf ## vhosts.d/vhosts.template ## var.vhosts_dir = server_root + "/vhosts" ## ## Cache for mod_compress ## ## used in: ## conf.d/compress.conf ## var.cache_dir = "/var/cache/lighttpd" ## ## Base directory for sockets. ## ## used in: ## conf.d/fastcgi.conf ## conf.d/scgi.conf ## var.socket_dir = home_dir + "/sockets" ## ####################################################################### ####################################################################### ## ## Load the modules. include "modules.conf" ## ####################################################################### ####################################################################### ## ## Basic Configuration ## --------------------- ## server.port = 80 ## ## Use IPv6? ## #server.use-ipv6 = "enable" ## ## bind to a specific IP ## #server.bind = "localhost" ## ## Run as a different username/groupname. ## This requires root permissions during startup. ## server.username = "lighttpd" server.groupname = "lighttpd" ## ## enable core files. ## #server.core-files = "disable" ## ## Document root ## server.document-root = server_root + "/lighttpd" ## ## The value for the "Server:" response field. ## ## It would be nice to keep it at "lighttpd". ## #server.tag = "lighttpd" ## ## store a pid file ## server.pid-file = state_dir + "/lighttpd.pid" ## ####################################################################### ####################################################################### ## ## Logging Options ## ------------------ ## ## all logging options can be overwritten per vhost. ## ## Path to the error log file ## server.errorlog = log_root + "/error.log" ## ## If you want to log to syslog you have to unset the ## server.errorlog setting and uncomment the next line. ## #server.errorlog-use-syslog = "enable" ## ## Access log config ## include "conf.d/access_log.conf" ## ## The debug options are moved into their own file. ## see conf.d/debug.conf for various options for request debugging. ## include "conf.d/debug.conf" ## ####################################################################### ####################################################################### ## ## Tuning/Performance ## -------------------- ## ## corresponding documentation: ## http://www.lighttpd.net/documentation/performance.html ## ## set the event-handler (read the performance section in the manual) ## ## possible options on linux are: ## ## select ## poll ## linux-sysepoll ## ## linux-sysepoll is recommended on kernel 2.6. ## server.event-handler = "linux-sysepoll" ## ## The basic network interface for all platforms at the syscalls read() ## and write(). Every modern OS provides its own syscall to help network ## servers transfer files as fast as possible ## ## linux-sendfile - is recommended for small files. ## writev - is recommended for sending many large files ## server.network-backend = "linux-sendfile" ## ## As lighttpd is a single-threaded server, its main resource limit is ## the number of file descriptors, which is set to 1024 by default (on ## most systems). ## ## If you are running a high-traffic site you might want to increase this ## limit by setting server.max-fds. ## ## Changing this setting requires root permissions on startup. see ## server.username/server.groupname. ## ## By default lighttpd would not change the operation system default. ## But setting it to 2048 is a better default for busy servers. ## ## With SELinux enabled, this is denied by default and needs to be allowed ## by running the following once : setsebool -P httpd_setrlimit on server.max-fds = 2048 ## ## Stat() call caching. ## ## lighttpd can utilize FAM/Gamin to cache stat call. ## ## possible values are: ## disable, simple or fam. ## server.stat-cache-engine = "simple" ## ## Fine tuning for the request handling ## ## max-connections == max-fds/2 (maybe /3) ## means the other file handles are used for fastcgi/files ## server.max-connections = 1024 ## ## How many seconds to keep a keep-alive connection open, ## until we consider it idle. ## ## Default: 5 ## #server.max-keep-alive-idle = 5 ## ## How many keep-alive requests until closing the connection. ## ## Default: 16 ## #server.max-keep-alive-requests = 18 ## ## Maximum size of a request in kilobytes. ## By default it is unlimited (0). ## ## Uploads to your server cant be larger than this value. ## #server.max-request-size = 0 ## ## Time to read from a socket before we consider it idle. ## ## Default: 60 ## #server.max-read-idle = 60 ## ## Time to write to a socket before we consider it idle. ## ## Default: 360 ## #server.max-write-idle = 360 ## ## Traffic Shaping ## ----------------- ## ## see /usr/share/doc/lighttpd/traffic-shaping.txt ## ## Values are in kilobyte per second. ## ## Keep in mind that a limit below 32kB/s might actually limit the ## traffic to 32kB/s. This is caused by the size of the TCP send ## buffer. ## ## per server: ## #server.kbytes-per-second = 128 ## ## per connection: ## #connection.kbytes-per-second = 32 ## ####################################################################### ####################################################################### ## ## Filename/File handling ## ------------------------ ## ## files to check for if .../ is requested ## index-file.names = ( "index.php", "index.rb", "index.html", ## "index.htm", "default.htm" ) ## index-file.names += ( "index.xhtml", "index.html", "index.htm", "default.htm", "index.php" ) ## ## deny access the file-extensions ## ## ~ is for backupfiles from vi, emacs, joe, ... ## .inc is often used for code includes which should in general not be part ## of the document-root url.access-deny = ( "~", ".inc" ) ## ## disable range requests for pdf files ## workaround for a bug in the Acrobat Reader plugin. ## $HTTP["url"] =~ "\.pdf$" { server.range-requests = "disable" } ## ## url handling modules (rewrite, redirect) ## #url.rewrite = ( "^/$" => "/server-status" ) #url.redirect = ( "^/wishlist/(.+)" => "http://www.example.com/$1" ) ## ## both rewrite/redirect support back reference to regex conditional using %n ## #$HTTP["host"] =~ "^www\.(.*)" { # url.redirect = ( "^/(.*)" => "http://%1/$1" ) #} ## ## which extensions should not be handle via static-file transfer ## ## .php, .pl, .fcgi are most often handled by mod_fastcgi or mod_cgi ## static-file.exclude-extensions = ( ".php", ".pl", ".fcgi", ".scgi" ) ## ## error-handler for status 404 ## #server.error-handler-404 = "/error-handler.html" #server.error-handler-404 = "/error-handler.php" ## ## Format: <errorfile-prefix><status-code>.html ## -> ..../status-404.html for 'File not found' ## #server.errorfile-prefix = "/srv/www/htdocs/errors/status-" ## ## mimetype mapping ## include "conf.d/mime.conf" ## ## directory listing configuration ## include "conf.d/dirlisting.conf" ## ## Should lighttpd follow symlinks? ## server.follow-symlink = "enable" ## ## force all filenames to be lowercase? ## #server.force-lowercase-filenames = "disable" ## ## defaults to /var/tmp as we assume it is a local harddisk ## server.upload-dirs = ( "/var/tmp" ) ## ####################################################################### ####################################################################### ## ## SSL Support ## ------------- ## ## To enable SSL for the whole server you have to provide a valid ## certificate and have to enable the SSL engine.:: ## ## ssl.engine = "enable" ## ssl.pemfile = "/path/to/server.pem" ## ## The HTTPS protocol does not allow you to use name-based virtual ## hosting with SSL. If you want to run multiple SSL servers with ## one lighttpd instance you must use IP-based virtual hosting: :: ## ## $SERVER["socket"] == "10.0.0.1:443" { ## ssl.engine = "enable" ## ssl.pemfile = "/etc/ssl/private/www.example.com.pem" ## server.name = "www.example.com" ## ## server.document-root = "/srv/www/vhosts/example.com/www/" ## } ## ## If you have a .crt and a .key file, cat them together into a ## single PEM file: ## $ cat /etc/ssl/private/lighttpd.key /etc/ssl/certs/lighttpd.crt \ ## > /etc/ssl/private/lighttpd.pem ## #ssl.pemfile = "/etc/ssl/private/lighttpd.pem" ## ## optionally pass the CA certificate here. ## ## #ssl.ca-file = "" ## ####################################################################### ####################################################################### ## ## custom includes like vhosts. ## #include "conf.d/config.conf" #include_shell "cat /etc/lighttpd/vhosts.d/*.conf" ## ####################################################################### ####################################################################### ### Custom Added by me #url.rewrite-once = (".*\.(js|ico|gif|jpg|png|css|jar|class)$" => "$0", "" => "/index.php") url.rewrite-once = ( ".*\?(.*)$" => "/index.php?$1", "^/js/.*$" => "$0", "^.*\.(js|ico|gif|jpg|png|css|swf |jar|class)$" => "$0", "" => "/index.php" ) # expire.url = ( "" => "access 1 days" ) include "myvhost-vhosts.conf" ####################################################################### Here is my Vhost file for lighttpd $HTTP["host"] =~ "192.168.8.35$" { server.document-root = "/var/www/lighttpd/qc41022012/public" server.errorlog = "/var/log/lighttpd/error.log" accesslog.filename = "/var/log/lighttpd/access.log" server.error-handler-404 = "/e404.php" } and here is my nginx.conf file user nginx; worker_processes 5; error_log /var/log/nginx/error.log warn; pid /var/run/nginx.pid; events { worker_connections 1024; } http { include /etc/nginx/mime.types; default_type application/octet-stream; log_format main '$remote_addr - $remote_user [$time_local] "$request" ' '$status $body_bytes_sent "$http_referer" ' '"$http_user_agent" "$http_x_forwarded_for"'; access_log /var/log/nginx/testsite/logs/access.log main; sendfile on; #tcp_nopush on; keepalive_timeout 65; #gzip on; # include /etc/nginx/conf.d/*.conf; ## I added this ## include /etc/nginx/sites-available/*; } Here is my NginX Vhost file server { server_name 192.168.8.91; access_log /var/log/nginx/myapps/logs/access.log; error_log /var/log/nginx/myapps/logs/error.log; root /var/www/html/myapps/public; location / { index index.html index.htm index.php; } location = /favicon.ico { return 204; access_log off; log_not_found off; } # location ~ \.php$ { # try_files $uri /index.php; # include /etc/nginx/fastcgi_params; # fastcgi_pass 127.0.0.1:9000; # fastcgi_index index.php; # fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name; # fastcgi_param SCRIPT_NAME $fastcgi_script_name; location ~ \.php.*$ { rewrite ^(.*.php)/ $1 last; fastcgi_pass 127.0.0.1:9000; fastcgi_index index.php; include fastcgi_params; fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name; # fastcgi_intercept_errors on; # fastcgi_param SCRIPT_FILENAME $document_root/index.php; # fastcgi_param PATH_INFO $uri; # fastcgi_pass 127.0.0.1:9000; # include fastcgi_params; } } We have a custom apps that we created that works great with lighttpd. I went through some headache also when we were trying to figure out how to make it work with lighttpd. this is the line that helps make it work in lighttpd. url.rewrite-once = ( ".*\?(.*)$" => "/index.php?$1", "^/js/.*$" => "$0", "^.*\.(js|ico|gif|jpg|png|css|swf |jar|class)$" => "$0", "" => "/index.php" ) but I couldn't figure out how to make it works in NginX. The webserver run just fine when we use the phpinfo.php test file. However as soon as I point it to my apps, nothing comes up. Check the error.log file and there's no error. Very mind boggling. I spent over 1 week trying to figure it out with no luck.. Please help?

    Read the article

  • Do I have an efficient APC Setup?

    - by Gaia
    Regarding my particular APC setup: APC 3.1.9 PHP 5.3.3 via fCGI Apache 2.2.15 CentOS 6.3 1) Is it setup properly to minimize overall memory usage? /etc/php.d/apc.ini has only one line: "apc.cache_by_default=0" Each domain for which I want to turn on APC has all the apropriate APC configurations in its own php.ini. 2) I would like to keep only one copy of apc.php that can be accessed via any of the vhosts on the server. What's the recommended way to do this? It seems that apc.php doesn't play well with apache Alias directive. apc.php only exists is on one of the vhosts, is set to 644 and it doesn't seem to matter who owns it: if I try to access it via an alias I get only gibberish.

    Read the article

  • Session timeout is very short on IIS 7.5

    - by Mehdi Mousavi
    I have a website on windows server 2008 and iis 7 on a VPS. It works fine and has no problems, but after moving it to a shared hosting server with IIS 7.5, the session is lost after 4 or 5 clicks (like 30 secs) and I have to login again and again. The two sites are same, I copied the site from the VPS exactly as-is to the shared hosting server. The session timeouts in both web.config files are same. On the shared hosting server I don't have access to IIS manager to manipulate the settings. All I have is Plesk Control Panel 9.5 and the website's web.config file. What could be causing this to happen?

    Read the article

  • printable PHP manual - 'all but the Function Reference section'

    - by JW01
    My Motivation I find it easier to learn things by reading 'offline'. I'd like to lean back and read the narrative part of a paper version of the official php manual. My Scuppered Plan My plan was to download the manual, print all but the Function Reference section and then read it. I have downloaded the "Single HTML file" version of the manual from the php.net download page. (That version did not contain any images, so I patched-in the ones from the Many HTML files version with no problem.) My plan was to open that "Single HTML file" in an HTML editor, delete the Function Reference section then print it out. Unfortunately, although I have tried three different editors, I have not been able to successfully load-up that massive html file to be able to edit it. Its about (~40MB). I started to look into the phpdoc framework with a view to rendering my own html docs from the source...but that's a steep learning curve for a newby..and is a last resort. I would use a file splitter, but they tend to split files crudely with no regard for html/xml/xhtml sematics. So the question is... Does anyone know know where you can download the php manual in a version that is a kind of half-way house between the 'Single HTML file' and the 'Many HTML files'? Ideally with the docs split into 3 parts: File 1 - stuff before the function reference File 2 - function reference File 3 - stuff after the function reference Or Can you suggest any editors/tools will enable me to split up this file myself?

    Read the article

  • PHP Aspect Oriented Design

    - by Devin Dixon
    This is a continuation of this Code Review question. What was taken away from that post, and other aspect oriented design is it is hard to debug. To counter that, I implemented the ability to turn tracing of the design patterns on. Turning trace on works like: //This can be added anywhere in the code Run::setAdapterTrace(true); Run::setFilterTrace(true); Run::setObserverTrace(true); //Execute the functon echo Run::goForARun(8); In the actual log with the trace turned on, it outputs like so: adapter 2012-02-12 21:46:19 {"type":"closure","object":"static","call_class":"\/public_html\/examples\/design\/ClosureDesigns.php","class":"Run","method":"goForARun","call_method":"goForARun","trace":"Run::goForARun","start_line":68,"end_line":70} filter 2012-02-12 22:05:15 {"type":"closure","event":"return","object":"static","class":"run_filter","method":"\/home\/prodigyview\/public_html\/examples\/design\/ClosureDesigns.php","trace":"Run::goForARun","start_line":51,"end_line":58} observer 2012-02-12 22:05:15 {"type":"closure","object":"static","class":"run_observer","method":"\/home\/prodigyview\/public_html\/public\/examples\/design\/ClosureDesigns.php","trace":"Run::goForARun","start_line":61,"end_line":63} When the information is broken down, the data translates to: Called by an adapter or filter or observer The function called was a closure The location of the closure Class:method the adapter was implemented on The Trace of where the method was called from Start Line and End Line The code has been proven to work in production environments and features various examples of to implement, so the proof of concept is there. It is not DI and accomplishes things that DI cannot. I wouldn't call the code boilerplate but I would call it bloated. In summary, the weaknesses are bloated code and a learning curve in exchange for aspect oriented functionality. Beyond the normal fear of something new and different, what are other weakness in this implementation of aspect oriented design, if any? PS: More examples of AOP here: https://github.com/ProdigyView/ProdigyView/tree/master/examples/design

    Read the article

  • Purely technical reasons for PHP as a first choice?

    - by JoelFan
    I know this may come off as a flame-y / troll-y, but I hope you will take my word for it that it's not my intention. I am just trying to understand the PHP phenomenon. After looking at the many technical issues with the language design of PHP, I am hard pressed to find any redeeming technical advantages where PHP surpasses all other languages. Before coming to the conclusion that there would simply be no reason to choose PHP as a development language on purely technical grounds, I would like to ask, if all non-technical factors were equal (such as what language the developers already know, what languages the hosting provider offers, language of existing code, cost, license, corporate fiat, etc.), would there be any type of new software system that would indicate making PHP a first choice for development? If so, what technical advantage does PHP have over all other languages that would cause you to choose it? EDIT: I am not interested in comparing PHP "out of the box" with other languages "out of the box". If PHP has a certain feature "out of the box" that another language has only after installing some readily available add-on, that is not considered an advantage for PHP for the purposes of this question.

    Read the article

  • Log php errors in ubuntu

    - by resting
    I followed the setup here: Where is the PHP error log When I look into /var/log/php_errors.log, I could see some PHP errors. PHP Warning: file_get_contents(/var/www/...): failed to open stream: No such file or directory in ... But what I'm trying to see is the error when I removed a semicolon from a statement. That error above has no relation to file from where I removed the semicolon so we can just ignore that. When I access the page with the removed semicolon, I get The website encountered an error while retrieving https://myapp/download/decode/testfile. It may be down for maintenance or configured incorrectly. HTTP Error 500 (Internal Server Error): An unexpected condition was encountered while the server was attempting to fulfill the request. But no logs in /var/log/php_errors.log. How do I see the error that usually says which line and which file the process failed? The real reason for trying to see the error is because I have a very huge loop, that throws the HTTP 500 error and I can't see the exact error. I'm just simulation with a removed semicolon to test things out. Other settings: error_reporting = E_ALL & ~E_DEPRECATED display_errors = On On Ubuntu 10.04.4 LTS Update Ok, I managed to get the error message to display. Parse error: syntax error, unexpected T_IF in ... However, it's still not logged. It wasn't displaying previously because Cakephp's debug level was at 0. Setting it to 2 displays the message, but no logs.

    Read the article

  • Suggestions for a CMS markup language for PHP

    - by Yanick Rochon
    As a learning experience, and as project, I am attempting to write a CMS module for ZF2. One of the functionality I would like to have is the possibility of adding dynamic contents in the pages by calling PHP functions in the view scripts. However, I do not want to give the users freedom in writing PHP code directly inside the page content, but rather implement custom view helpers (or widgets) to handle logic. For example: calling partial, partialLoop, url, etc. specifying arguments and all. I liked the idea of extending Markdown but this would get complicated when trying to add custom CSS class to elements, etc. Then I had the idea of simply doing a preg_replace on some patterns. For example, the string : ### partialLoop:['partials/display.phtml',[{id:'p1',price:4.99},{id:'p2',price:12.34}]] ### would be replaced by <?php echo $this->partialLoop('partials/display.phtml', array(array('id'=>'p1','price'=>4.99),array('id'=>'p2','price'=>12.34))) ?> Obviously, there would be some caching done so the page content is not rendered everytime. Does this sound good? If not, what would be a good way of doing this? Or is there a project already being developed for doing this? (I'd like to avoid heavy third party libs and something fairly or fully compatible with ZF2 would be nice.) Thanks.

    Read the article

  • UTF-8 encoding problem with flash mysql and php

    - by alibhp
    Hi, As you may know, I am programming an on-line game using FLASH. I am connecting my FLASH 8 movie with MySQL database through PHP. I am doing very good in that, and I have everything working fine. The problems come when I am trying to insert (Using the INSERT SQL func) data to the database that are non-english. In other words, UTF-8 data. I red a lot of articls about that stuff and found and apply the fallowing: 1. In PHP4, you need to tell the PHP to use UTF-8 when using the xml_parser_crater() func, however, in PHP5 that is done automatically. Even though I told PHP5 to use the UTF-8 when calling the func. Adding the header to the XML sent to PHP from flash. Force the FLASH to use UTF-8 encoding in the preference options. Set the encoding in MySQL to UTF-8 (utf8_unicode_ci with InnoDB engine). I can read and insert the other language data correctly in the phpadmin as well. I did all that in my coding, and still I can't insert such data. one more strange thing is that, when I use the same link, that the FLASH using, with the XML, that the FLASH creating, on the browser (google chrome), I got the data inserted right in the database!!!!! I am about to get crazy about that stuff, What am I missing? what cause the problem? Thank you in advance.

    Read the article

  • phpize: m4 error just in one extension

    - by Francois
    On Linux, I installed php 5.3.8 from source. Using phpize for installing an extension works fine but not on one specific extension (mysqlnd). # cd /opt/php/5.3.8/ext/pdo && /opt/php/5.3.8/bin/phpize ... this runs ok # cd /opt/php/5.3.8/ext/mysqlnd && /opt/php/5.3.8/bin/phpize Cannot find config.m4. Make sure that you run '/opt/php/5.3.8/bin/phpize' in the top level source directory of the module` As you can see error can not be that I am not on top level source directory since I am. I tried to call phpize from ext folder - did not work either! For info I have m4 installed Any idea? Thanks :)

    Read the article

  • How-to enable user session time out warning (JDev 11.1.1.4)

    - by frank.nimphius
    Normal 0 false false false EN-US X-NONE X-NONE /* Style Definitions */ table.MsoNormalTable {mso-style-name:"Table Normal"; mso-tstyle-rowband-size:0; mso-tstyle-colband-size:0; mso-style-noshow:yes; mso-style-priority:99; mso-style-qformat:yes; mso-style-parent:""; mso-padding-alt:0in 5.4pt 0in 5.4pt; mso-para-margin:0in; mso-para-margin-bottom:.0001pt; mso-pagination:widow-orphan; font-size:11.0pt; font-family:"Calibri","sans-serif"; mso-ascii-font-family:Calibri; mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman"; mso-fareast-theme-font:minor-fareast; mso-hansi-font-family:Calibri; mso-hansi-theme-font:minor-latin; mso-bidi-font-family:"Times New Roman"; mso-bidi-theme-font:minor-bidi;} Oracle JDeveloper 11.1.1.4 contains a new session time-out warning functionality. Quoting the Oracle® Fusion Middleware Web User Interface Developer's Guide for Oracle Application Development Framework11g Release 1 (11.1.1.4.0) documentatiom http://download.oracle.com/docs/cd/E17904_01/web.1111/b31973/ap_config.htm#BABFIGBA "When a request is sent to the server, a session timeout value is written to the page and the session timeout warning interval is defined by the context parameter  oracle.adf.view.rich.sessionHandling.WARNING_BEFORE_TIMEOUT. The user is given the opportunity to extend the session in a warning dialog, and a notification is sent when the session has expired and the page is refreshed. Depending on the application security configuration, the user may be redirected to the log in page when the session expires. Use the oracle.adf.view.rich.sessionHandling.WARNING_BEFORE_TIMEOUT context parameter to set the number of seconds prior to the session time out when a warning dialog is displayed. If the value of WARNING_BEFORE_TIMEOUT is less than 120 seconds, if client state saving is used for the page, or if the session has been invalidated, the feature is disabled. The session time-out value it taken directly from the session. Example A-3 shows configuration of the warning dialog to display at 120 seconds before the time-out of the session. Example A-3 Configuration of Session Time-out Warning <context-param>    <param-name>        oracle.adf.view.rich.sessionHandling.WARNING_BEFORE_TIMEOUT   </param-name>    <param-value>120</param-value> </context-param> The default value of this parameter is 120 seconds. To prevent notification of the user too frequently when the session time-out is set too short, the actual value of WARNING_BEFORE_TIMEOUT is determined dynamically, where the session time-out must be more than 2 minutes or the feature is disabled.

    Read the article

  • PHP MYSQL loop to check if LicenseID Values are contained in mysql DB [closed]

    - by Jasper
    I have some troubles to find the right loop to check if some values are contained in mysql DB. I'm making a software and I want to add license ID. Each user has x keys to use. Now when the user start the client, it invokes a PHP page that check if the Key sent in the POST method is stored in DB or not. If that key isn't store than I need to check the number of his keys. If it's than X I'll ban him otherwise i add the new keys in the DB. I'm new with PHP and MYSQL. I wrote this code and I would know if I can improve it. <?php $user = POST METHOD $licenseID = POST METHOD $resultLic= mysql_query("SELECT id , idUser , idLicense FROM license WHERE idUser = '$user'") or die(mysql_error()); $resultNumber = mysql_num_rows($resultLic); $keyFound = '0'; // If keyfound is 1 the key is stored in DB while ($rows = mysql_fetch_array($resultLic,MYSQL_BOTH)) { //this loop check if the $licenseID is stored in DB or not for($i=0; $i< $resultNumber ; i++) { if($rows['idLicense'] === $licenseID) { //Just for the debug echo("License Found"); $keyFound = '1'; break; } //If key isn't in DB and there are less than 3 keys the new key will be store in DB if($keyfound == '0' && $resultNumber < 3) { mysql_query( Update users set ...Store $licenseID in Table) } // Else mean that the user want user another generated key (from the client) in the DB and i will be ban (It's wrote in TOS terms that they cant use the software on more than 3 different station) else { mysql_query( update users set ban ='1'.....etc ); } } ?> I know that this code seems really bad so i would know how i can improve it. Someone Could give me any advice? I choose to have 2 tables: users where all information about the users is, with fields id, username, password and another table license with fields id, idUsername, idLicense (the last one store license that the software generate)

    Read the article

  • Choosing between PHP and Java

    - by user996459
    I've recently started University, studying Computing and IT. My Uni focuses on Java. My study will consist of mathematics, 'boring' IT related stuff and several Java units such as: -Software development with Java, -Object-oriented Java programming, -Relational databases: theory and practice (using Java), -Developing concurrent distributed systems (using Java), -Software engineering with objects (using Java). I'm trying to determine whenever I should focus on Java and self study it in my free time so that I can actually learn and become a competent Java programmer by the time I graduate, or, only do enough Java to get the degree but in my free time self study PHP and related web technologies. Job market in my area appears to be balanced for the two, salary and availability wise. Regardless of which patch I'd take getting a job should not be a problem however Java does seem to pay almost insignificantly more. In terms of my interest and career expectations, I don't have anything specific planned. I very much enjoy writing code but I don't really care what kind. So far I equally enjoyed writing C, AutoIT, vb.net, PHP and even Java. Basically, I'm happy as long as I get to type in code (be it low level programming or web back-end scripting). So the question really is, would my Uni and their Java focus profit me should I choose PHP? Or should I buy what my university is selling and stick to Java like a fly sticks to poop...? Apologies for cryptic writing, still learning English

    Read the article

  • IE session (-nomerge) management application?

    - by skrco
    I'm in need of an application that can manage multiple Internet Explorer instances (to be precise nomerge sessions) like you can in Remote Desktop Manager with RDPs. This app should host them in single window and arrange IE instances e.g. in tabs or lists. OK, in Remote Desktop Manager you can create Web session, but in embedded mode you cannot set the nomerge option - all windows and tabs share the same session. I've been searching the web, but with no results. So I put this question whether anyone know of such application or any workaround. Or I have to write my own app?

    Read the article

  • Either, nginx+php-fpm bad config or nginx+php-fpm cannot handle high query?

    - by The Wolf
    I have wordpress installed in my server configured(hopefully with nginx+php-fpm+mariaDB). I am trying to import using wordpress importer a 1.5MB xml file. Everytime I try to upload it using the importer, it got cut of... meaning just blank screen result.. Here is my error log: actually I just posted 2 of the errors [error] 858#0: *1 connect() failed (111: Connection refused) while connecting to upstream, client: xx.xxx.xx.xx, server: xxx.com, request: "GET xxxx.html HTTP/1.1", upstream: "fastcgi://127.0.0.1:9000", host: "xxx.com" [error] 858#0: *13 connect() failed (111: Connection refused) while connecting to upstream, client: xxx.x.xx.xx, server: xxx.com, request: "GET xxxx.php HTTP/1.1", upstream: "fastcgi://127.0.0.1:9000", host: "xxx.com" I don't know what is the reason why it can't process the wordpress export .xml. I already increased max_file_upload & etc., but nothing happens. Hope somebody can help me. Here are my conf: nginx.conf user nginx; worker_processes 8; error_log /var/log/nginx/error.log warn; pid /var/run/nginx.pid; events { worker_connections 1024; } http { include /etc/nginx/mime.types; default_type application/octet-stream; log_format main '$remote_addr - $remote_user [$time_local] "$request" ' '$status $body_bytes_sent "$http_referer" ' '"$http_user_agent" "$http_x_forwarded_for"'; access_log /var/log/nginx/access.log main; sendfile on; #tcp_nopush on; server_tokens off; keepalive_timeout 65; fastcgi_read_timeout 500; #gzip on; client_max_body_size 2M; php-fpm.conf ;;;;;;;;;;;;;;;;;;;;; ; FPM Configuration ; ;;;;;;;;;;;;;;;;;;;;; ; All relative paths in this configuration file are relative to PHP's install ; prefix. ; Include one or more files. If glob(3) exists, it is used to include a bunch of ; files from a glob(3) pattern. This directive can be used everywhere in the ; file. include=/etc/php-fpm.d/*.conf ;;;;;;;;;;;;;;;;;; ; Global Options ; ;;;;;;;;;;;;;;;;;; [global] ; Pid file ; Default Value: none pid = /var/run/php-fpm/php-fpm.pid ; Error log file ; Default Value: /var/log/php-fpm.log error_log = /var/log/php-fpm/error.log ; Log level ; Possible Values: alert, error, warning, notice, debug ; Default Value: notice ;log_level = notice ; If this number of child processes exit with SIGSEGV or SIGBUS within the time ; interval set by emergency_restart_interval then FPM will restart. A value ; of '0' means 'Off'. ; Default Value: 0 ;emergency_restart_threshold = 0 ; Interval of time used by emergency_restart_interval to determine when ; a graceful restart will be initiated. This can be useful to work around ; accidental corruptions in an accelerator's shared memory. ; Available Units: s(econds), m(inutes), h(ours), or d(ays) ; Default Unit: seconds ; Default Value: 0 ;emergency_restart_interval = 0 ; Time limit for child processes to wait for a reaction on signals from master. ; Available units: s(econds), m(inutes), h(ours), or d(ays) ; Default Unit: seconds ; Default Value: 0 ;process_control_timeout = 0 ; Send FPM to background. Set to 'no' to keep FPM in foreground for debugging. ; Default Value: yes daemonize = no ;;;;;;;;;;;;;;;;;;;; ; Pool Definitions ; ;;;;;;;;;;;;;;;;;;;; ; See /etc/php-fpm.d/*.conf [root@host etc]# vim php-fpm.conf [root@host etc]# vim php-fpm.conf ; Default Value: notice ;log_level = notice ; If this number of child processes exit with SIGSEGV or SIGBUS within the time ; interval set by emergency_restart_interval then FPM will restart. A value ; of '0' means 'Off'. ; Default Value: 0 ;emergency_restart_threshold = 0 ; Interval of time used by emergency_restart_interval to determine when ; a graceful restart will be initiated. This can be useful to work around ; accidental corruptions in an accelerator's shared memory. ; Available Units: s(econds), m(inutes), h(ours), or d(ays) ; Default Unit: seconds ; Default Value: 0 ;emergency_restart_interval = 0 ; Time limit for child processes to wait for a reaction on signals from master. ; Available units: s(econds), m(inutes), h(ours), or d(ays) ; Default Unit: seconds ; Default Value: 0 ;process_control_timeout = 0 ; Send FPM to background. Set to 'no' to keep FPM in foreground for debugging. ; Default Value: yes daemonize = no ;;;;;;;;;;;;;;;;;;;; ; Pool Definitions ; ;;;;;;;;;;;;;;;;;;;; ; See /etc/php-fpm.d/*.conf ps aux [root@host etc]# ps aux USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND root 1 0.0 0.1 2900 1380 ? Ss Jun02 0:00 init root 2 0.0 0.0 0 0 ? S Jun02 0:00 [kthreadd/9308] root 3 0.0 0.0 0 0 ? S Jun02 0:00 [khelper/9308] root 124 0.0 0.0 2464 576 ? S<s Jun02 0:00 /sbin/udevd -d root 460 0.0 0.1 35976 1308 ? Sl Jun02 0:00 /sbin/rsyslogd -i /var/run/syslogd.pid -c 5 root 474 0.0 0.0 8940 1028 ? Ss Jun02 0:00 /usr/sbin/sshd root 481 0.0 0.0 3264 876 ? Ss Jun02 0:00 xinetd -stayalive -pidfile /var/run/xinetd.pid root 491 0.0 0.1 6268 1432 ? S Jun02 0:00 /bin/sh /usr/bin/mysqld_safe --datadir=/var/lib/mysql --pid-file=/var/lib/mysql/host.busilak.com. mysql 584 0.1 6.8 679072 71456 ? Sl Jun02 0:04 /usr/sbin/mysqld --basedir=/usr --datadir=/var/lib/mysql --plugin-dir=/usr/lib/mysql/plugin --use root 586 0.0 0.3 12008 3820 ? Ss Jun02 0:01 sshd: root@pts/0 root 629 0.0 0.0 9140 756 ? Ss Jun02 0:00 /usr/sbin/saslauthd -m /var/run/saslauthd -a pam -n 2 root 630 0.0 0.0 9140 520 ? S Jun02 0:00 /usr/sbin/saslauthd -m /var/run/saslauthd -a pam -n 2 root 645 0.0 0.1 12788 1928 ? Ss Jun02 0:01 sendmail: accepting connections smmsp 653 0.0 0.1 12576 1728 ? Ss Jun02 0:00 sendmail: Queue runner@01:00:00 for /var/spool/clientmqueue root 691 0.0 0.1 7148 1184 ? Ss Jun02 0:00 crond root 698 0.0 0.1 6272 1688 pts/0 Ss Jun02 0:00 -bash root 1006 0.0 0.0 7828 924 ? Ss 00:30 0:00 nginx: master process /usr/sbin/nginx -c /etc/nginx/nginx.conf nginx 1007 0.0 0.1 8156 1724 ? S 00:30 0:00 nginx: worker process nginx 1008 0.0 0.1 8024 1360 ? S 00:30 0:00 nginx: worker process nginx 1009 0.0 0.1 8020 1356 ? S 00:30 0:00 nginx: worker process nginx 1011 0.0 0.1 8024 1360 ? S 00:30 0:00 nginx: worker process nginx 1012 0.0 0.1 8024 1360 ? S 00:30 0:00 nginx: worker process nginx 1013 0.0 0.1 8024 1360 ? S 00:30 0:00 nginx: worker process nginx 1014 0.0 0.1 8024 1360 ? S 00:30 0:00 nginx: worker process nginx 1015 0.0 0.1 8024 1344 ? S 00:30 0:00 nginx: worker process root 1030 0.0 0.2 25396 2904 ? Ss 00:30 0:00 php-fpm: master process (/etc/php-fpm.conf) apache 1031 0.0 1.9 40700 20624 ? S 00:30 0:00 php-fpm: pool www apache 1032 0.0 2.0 41924 21888 ? S 00:30 0:01 php-fpm: pool www apache 1033 0.0 1.9 41212 20848 ? S 00:30 0:01 php-fpm: pool www apache 1034 0.0 1.9 40956 20792 ? S 00:30 0:01 php-fpm: pool www apache 1035 0.0 2.0 41560 21556 ? S 00:30 0:02 php-fpm: pool www apache 1040 0.0 1.8 39292 19120 ? S 00:30 0:00 php-fpm: pool www root 1125 0.0 0.0 6080 1040 pts/0 R+ 01:04 0:00 ps aux netstat -l [root@host etc]# netstat -l Active Internet connections (only servers) Proto Recv-Q Send-Q Local Address Foreign Address State tcp 0 0 *:ssh *:* LISTEN tcp 0 0 localhost.localdomain:smtp *:* LISTEN tcp 0 0 localhost.locald:cslistener *:* LISTEN tcp 0 0 *:mysql *:* LISTEN tcp 0 0 *:http *:* LISTEN tcp 0 0 *:ssh *:* LISTEN Active UNIX domain sockets (only servers) Proto RefCnt Flags Type State I-Node Path unix 2 [ ACC ] STREAM LISTENING 60575947 /var/run/saslauthd/mux unix 2 [ ACC ] STREAM LISTENING 60574168 @/com/ubuntu/upstart unix 2 [ ACC ] STREAM LISTENING 60575873 /var/lib/mysql/mysql.sock Hope somebody can help me to figure out what is the problem.

    Read the article

  • Custom Session Management using HashTable

    - by kaleidoscope
    ASP.NET session state lets you associate a server-side string or object dictionary containing state data with a particular HTTP client session. A session is defined as a series of requests issued by the same client within a certain period of time, and is managed by associating a session ID with each unique client. The ID is supplied by the client on each request, either in a cookie or as a special fragment of the request URL. The session data is stored on the server side in one of the supported session state stores, which include in-process memory, SQL Server™ database, and the ASP.NET State Server service. The latter two modes enable session state to be shared among multiple Web servers on a Web farm and do not require server affinity. Implement Custom session Handler you need to follow following process : 1. Create class library which will inherit from  SessionStateStoreProviderBase abstract Class. 2. Implement all abstract Method in your base class. 3.Change Mode of session to “Custom” in web.config file and provide Provider as your Namespace with classname. <sessionState mode=”Custom” customProvider=”Namespace.classname”> <Providers> <add name=”Name” type=”Namespace.classname”> </sessionstate> For more Details Please refer following links :   http://msdn.microsoft.com/en-us/magazine/cc163730.aspx http://msdn.microsoft.com/en-us/library/system.web.sessionstate.sessionstatestoreproviderbase.aspx - Chandraprakash, S Technorati Tags: Chandraprakash,Session state Managment

    Read the article

  • OOP oriented PHP app source code samples and advice

    - by abel
    The day I have been dreading has arrived. I never felt OOP or good software design was important(I knew they were important, but I thought I could manage without them.). However having read otherwise almost everywhere on the interwebs, I started dreading the day when my client would ask me for new features in an existing app. The day has come and the pain is unbearable! I have never coded my PHP websites "properly"(PHP is my primary language and the bulk of my work. I am learning Python (using web2py)) I take care that the website doesn't fall apart in a daily use scenario. I code pages like I was creating a list of static html files with bits of "magic code" in each of them(this bugs me a lot). How do I make the whole app more or less a single object? For eg. How do I design the object model for an invoicing app? I use a lot of functions for doing any particular thing in the same fashion throughout the app(for eg. validation, generating ids, calculating taxes etc.). I know the basics of OOP in general. Can anyone point me to source code samples of functional apps written in php? Or can someone provide pointers so I can recode my existing apps in a more modular way.

    Read the article

< Previous Page | 32 33 34 35 36 37 38 39 40 41 42 43  | Next Page >