Details

    • Type: Bug Bug
    • Status: Closed
    • Priority: Blocker Blocker
    • Resolution: Fixed
    • Affects Version/s: 2.1.0
    • Fix Version/s: 2.1.0
    • Component/s: None
    • Environment:

      Ubuntu 8.04, Linux 2.6.18-028stab067.4-PAE #1 SMP Thu Jan 14 22:09:19 MSK 2010 i686 GNU/Linux - Virtuozzo VPS

      Description

      Having compiled TS, and started the server, I get the following errors both in the syslog and in traffic.out (some of the messages in traffic.out are very similar to those in the syslog) :

      in the syslog :

      Mar 20 19:36:47 tagmata traffic_cop[1795]: — Cop Starting Version: Apache Traffic Server - traffic_cop - 2.1.0 - (build # 2204 on Mar 20 2010 at 04:09:00)
      Mar 20 19:36:47 tagmata traffic_cop[1795]: traffic_manager not running, making sure traffic_server is dead
      Mar 20 19:36:47 tagmata traffic_cop[1795]: spawning traffic_manager
      Mar 20 19:36:47 tagmata traffic_manager[1797]: NOTE: — Manager Starting —
      Mar 20 19:36:47 tagmata traffic_manager[1797]: NOTE: Manager Version: Apache Traffic Server - traffic_manager - 2.1.0 - (build # 2204 on Mar 20 2010 at 04:11:51)
      Mar 20 19:36:47 tagmata traffic_manager[1797]: NOTE: RLIMIT_NOFILE(7):cur(10000),max(10000)
      Mar 20 19:36:47 tagmata traffic_manager[1797]:

      {3079632576} STATUS: opened var/log/trafficserver/manager.log
      Mar 20 19:36:47 tagmata traffic_manager[1797]: {3079632576}

      NOTE: updated diags config
      Mar 20 19:36:47 tagmata traffic_manager[1797]:

      {3079632576} NOTE: [ClusterCom::ClusterCom] Node running on OS: 'Linux' Release: '2.6.18-028stab067.4-PAE'
      Mar 20 19:36:47 tagmata traffic_manager[1797]: {3079632576}

      NOTE: [LocalManager::listenForProxy] Listening on port: 8080
      Mar 20 19:36:47 tagmata traffic_manager[1797]:

      {3079632576} NOTE: [TrafficManager] Setup complete
      Mar 20 19:36:48 tagmata traffic_manager[1797]: {3079632576}

      NOTE: [LocalManager::startProxy] Launching ts process
      Mar 20 19:36:48 tagmata traffic_manager[1797]:

      {3079632576} NOTE: [LocalManager::pollMgmtProcessServer] New process connecting fd '14'
      Mar 20 19:36:48 tagmata traffic_manager[1797]: {3079632576}

      NOTE: [Alarms::signalAlarm] Server Process born
      Mar 20 19:36:49 tagmata traffic_server[1808]: NOTE: — Server Starting —
      Mar 20 19:36:49 tagmata traffic_server[1808]: NOTE: Server Version: Apache Traffic Server - traffic_server - 2.1.0 - (build # 2204 on Mar 20 2010 at 04:13:15)
      Mar 20 19:36:49 tagmata traffic_server[1808]:

      {1080614544} STATUS: opened var/log/trafficserver/diags.log
      Mar 20 19:36:49 tagmata traffic_server[1808]: {1080614544}

      NOTE: updated diags config
      Mar 20 19:36:49 tagmata traffic_server[1808]:

      {1080614544} NOTE: cache clustering disabled
      Mar 20 19:36:49 tagmata traffic_server[1808]: {1080614544}

      NOTE: clearing statistics
      Mar 20 19:36:49 tagmata traffic_server[1808]: FATAL: ../../../../trafficserver/iocore/eventsystem/UnixEThread.cc:70: failed assert `(evfd = eventfd(0,0)) >= 0`
      Mar 20 19:36:49 tagmata traffic_manager[1797]:

      {3079632576} ERROR: [LocalManager::pollMgmtProcessServer] Server Process terminated due to Sig 6: Aborted
      Mar 20 19:36:49 tagmata traffic_manager[1797]: {3079632576}

      ERROR: (last system error 2: No such file or directory)
      Mar 20 19:36:49 tagmata traffic_manager[1797]:

      {3079632576} ERROR: [Alarms::signalAlarm] Server Process was reset
      Mar 20 19:36:49 tagmata traffic_manager[1797]: {3079632576}

      ERROR: (last system error 2: No such file or directory)
      Mar 20 19:36:50 tagmata traffic_manager[1797]:

      {3079632576} NOTE: [LocalManager::startProxy] Launching ts process
      Mar 20 19:36:50 tagmata traffic_manager[1797]: {3079632576}

      NOTE: [LocalManager::pollMgmtProcessServer] New process connecting fd '14'
      Mar 20 19:36:50 tagmata traffic_manager[1797]:

      {3079632576}

      NOTE: [Alarms::signalAlarm] Server Process born

      in traffic.out :

      [Mar 20 20:02:22.778] Manager

      {3080124096} ERROR: [LocalManager::pollMgmtProcessServer] Server Process terminated due to Sig 6: Aborted
      [Mar 20 20:02:22.778] Manager {3080124096}

      ERROR: (last system error 2: No such file or directory)
      [Mar 20 20:02:22.778] Manager

      {3080124096} ERROR: [Alarms::signalAlarm] Server Process was reset
      [Mar 20 20:02:22.778] Manager {3080124096}

      ERROR: (last system error 2: No such file or directory)
      [TrafficServer] using root directory '/simpl/apps/trafficserver/trunk'
      [Mar 20 20:02:24.830]

      {1080614544}

      STATUS: opened var/log/trafficserver/diags.log
      FATAL: ../../../../trafficserver/iocore/eventsystem/UnixEThread.cc:70: failed assert `(evfd = eventfd(0,0)) >= 0`
      bin/traffic_server - STACK TRACE:
      bin/traffic_server(ink_fatal+0x1a)[0x834b6fa]
      bin/traffic_server(_ink_assert+0x68)[0x834a898]
      bin/traffic_server(_ZN7EThreadC1E10ThreadTypei+0x3f3)[0x8344b93]
      bin/traffic_server(_ZN14EventProcessor5startEi+0x8d)[0x8345f9d]
      bin/traffic_server(main+0xaf3)[0x8117533]
      /lib/libc.so.6(__libc_start_main+0xe0)[0x40554450]
      bin/traffic_server[0x80c3231]

      The same messages are then repeated every two seconds or so.

      I used the default apt-get install's for all the dependencies, and apart from Sqlite, which uses the latest source, these files were used in the build.

      The kernel is quite an old one now, but unfortunately I cannot upgrade because my ISP doesn't offer a more recent kernel.

      Although I can telnet into the server, it won't process any requests.

      1. records.config
        26 kB
        Marcus Clyne
      2. 0001-TS263_patch1.diff.patch
        1 kB
        George Paul
      3. Kernel 2.6.21.7-2 Errors log with Patch (p1).txt
        2 kB
        Sean Harmer
      4. TS-263__configure.ac.patch
        1 kB
        Jason Giedymin

        Issue Links

          Activity

          Hide
          George Paul added a comment -

          These fixes are only applicable to 2.1/trunk branch where eventfd() support was added and used as the default instead of default pipes usage in 2.0.x.
          -George

          Show
          George Paul added a comment - These fixes are only applicable to 2.1/trunk branch where eventfd() support was added and used as the default instead of default pipes usage in 2.0.x. -George
          Hide
          Sean Harmer added a comment -

          It's working!!!!!!!! Thanks to our resident ATS geniuses: Jason, George, and Leif

          Show
          Sean Harmer added a comment - It's working!!!!!!!! Thanks to our resident ATS geniuses: Jason, George, and Leif
          Hide
          Jason Giedymin added a comment -

          try (after patching):

          ./configure --disable-eventfd

          Show
          Jason Giedymin added a comment - try (after patching): ./configure --disable-eventfd
          Hide
          Sean Harmer added a comment -

          Installed the new patch (TS-263__configure.ac.patch)
          Rebuilt ATS on ec2 instance
          setup config files (storage, records, and remap)
          Pointed storage to root (' . ")
          Started ATS

          Here is the new /var/log messages/errors:

          Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server[28948]: NOTE: — Server Starting —
          Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server[28948]: NOTE: Server Version: Apache Traffic Server - traffic_server - 2.1.0 - (build # 22522 on Mar 25 2010 at 22:52:09)
          Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server[28948]:

          {1081907712} STATUS: opened var/log/trafficserver/diags.log
          Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server[28948]: {1081907712}

          NOTE: updated diags config
          Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server[28948]:

          {1081907712} NOTE: cache clustering disabled
          Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server[28948]: {1081907712}

          NOTE: clearing statistics
          Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server[28948]: FATAL: UnixEThread.cc:70: failed assert `(evfd = eventfd(0,0)) >= 0`
          Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_manager[28644]:

          {3086346592} ERROR: [LocalManager::pollMgmtProcessServer] Server Process terminated due to Sig 6: Aborted
          Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_manager[28644]: {3086346592}

          ERROR: (last system error 2: No such file or directory)
          Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_manager[28644]:

          {3086346592} ERROR: [Alarms::signalAlarm] Server Process was reset
          Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_manager[28644]: {3086346592}

          ERROR: (last system error 2: No such file or directory)
          Mar 25 23:06:56 domU-12-31-39-09-C4-83 traffic_manager[28644]:

          {3086346592} NOTE: [LocalManager::startProxy] Launching ts process
          Mar 25 23:06:56 domU-12-31-39-09-C4-83 traffic_manager[28644]: {3086346592}

          NOTE: [LocalManager::pollMgmtProcessServer] New process connecting fd '9'
          Mar 25 23:06:56 domU-12-31-39-09-C4-83 traffic_manager[28644]:

          {3086346592}

          NOTE: [Alarms::signalAlarm] Server Process born

          Any thoughts?

          Sorry to keep coming up with problem.

          Sean

          Show
          Sean Harmer added a comment - Installed the new patch ( TS-263 __configure.ac.patch) Rebuilt ATS on ec2 instance setup config files (storage, records, and remap) Pointed storage to root (' . ") Started ATS Here is the new /var/log messages/errors: Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server [28948] : NOTE: — Server Starting — Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server [28948] : NOTE: Server Version: Apache Traffic Server - traffic_server - 2.1.0 - (build # 22522 on Mar 25 2010 at 22:52:09) Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server [28948] : {1081907712} STATUS: opened var/log/trafficserver/diags.log Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server [28948] : {1081907712} NOTE: updated diags config Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server [28948] : {1081907712} NOTE: cache clustering disabled Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server [28948] : {1081907712} NOTE: clearing statistics Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_server [28948] : FATAL: UnixEThread.cc:70: failed assert `(evfd = eventfd(0,0)) >= 0` Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_manager [28644] : {3086346592} ERROR: [LocalManager::pollMgmtProcessServer] Server Process terminated due to Sig 6: Aborted Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_manager [28644] : {3086346592} ERROR: (last system error 2: No such file or directory) Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_manager [28644] : {3086346592} ERROR: [Alarms::signalAlarm] Server Process was reset Mar 25 23:06:55 domU-12-31-39-09-C4-83 traffic_manager [28644] : {3086346592} ERROR: (last system error 2: No such file or directory) Mar 25 23:06:56 domU-12-31-39-09-C4-83 traffic_manager [28644] : {3086346592} NOTE: [LocalManager::startProxy] Launching ts process Mar 25 23:06:56 domU-12-31-39-09-C4-83 traffic_manager [28644] : {3086346592} NOTE: [LocalManager::pollMgmtProcessServer] New process connecting fd '9' Mar 25 23:06:56 domU-12-31-39-09-C4-83 traffic_manager [28644] : {3086346592} NOTE: [Alarms::signalAlarm] Server Process born Any thoughts? Sorry to keep coming up with problem. Sean
          Hide
          Jason Giedymin added a comment -

          To use the patch i run:

          patch -p0 < TS-263__configure.ac.patch

          In the location where configure.ac is located.

          Show
          Jason Giedymin added a comment - To use the patch i run: patch -p0 < TS-263 __configure.ac.patch In the location where configure.ac is located.
          Hide
          Leif Hedstrom added a comment -

          Suggesting we fix this for the 2.0.0 release as well.

          Show
          Leif Hedstrom added a comment - Suggesting we fix this for the 2.0.0 release as well.
          Hide
          Jason Giedymin added a comment -

          New patch to disable eventfd.

          Show
          Jason Giedymin added a comment - New patch to disable eventfd.
          Hide
          Sean Harmer added a comment -

          Attached is the output of my /var/log errors, after installing the first patch (p1)

          Show
          Sean Harmer added a comment - Attached is the output of my /var/log errors, after installing the first patch (p1)
          Hide
          Jason Giedymin added a comment -

          New patch is coming, the current one doesn't work.

          Show
          Jason Giedymin added a comment - New patch is coming, the current one doesn't work.
          Hide
          Jason Giedymin added a comment -

          I'm testing this patch now.

          Show
          Jason Giedymin added a comment - I'm testing this patch now.
          Hide
          George Paul added a comment -

          Hi folks,

          Please try this patch '0001-TS263_patch1.diff.patch'

          patch -p1 -i 0001-TS263_patch1.diff.patch

          -George

          Show
          George Paul added a comment - Hi folks, Please try this patch '0001-TS263_patch1.diff.patch' patch -p1 -i 0001-TS263_patch1.diff.patch -George
          Hide
          George Paul added a comment -

          Jason,
          I agree that as driver developer what you suggested above is the way to go. The other way is to test for which eventfd() usage works in configure which is a bit more work. I'm okay if you go the first route but make a note of it, maybe in a comment.
          -George

          Show
          George Paul added a comment - Jason, I agree that as driver developer what you suggested above is the way to go. The other way is to test for which eventfd() usage works in configure which is a bit more work. I'm okay if you go the first route but make a note of it, maybe in a comment. -George
          Hide
          Jason Giedymin added a comment - - edited

          I sent this to the list, (though this is updated)

          My first instinct as a driver developer is to do:
          #if KERNEL_VERSION <= LINUX_VERSION_CODE(2,6,26)
          Use 2.6.2 API version of eventfd
          #else
          Use 2.6.27 API version of eventfd
          #endif
          ...

          Maybe wrap it, so it doesn't clutter the file up dunno

          Thoughts?

          Show
          Jason Giedymin added a comment - - edited I sent this to the list, (though this is updated) My first instinct as a driver developer is to do: #if KERNEL_VERSION <= LINUX_VERSION_CODE(2,6,26) Use 2.6.2 API version of eventfd #else Use 2.6.27 API version of eventfd #endif ... Maybe wrap it, so it doesn't clutter the file up dunno Thoughts?
          Hide
          Jason Giedymin added a comment -

          Eventfd is returning an error and that's why the assert follows through.

          George, I'll see what's behind this error tonight.

          Show
          Jason Giedymin added a comment - Eventfd is returning an error and that's why the assert follows through. George, I'll see what's behind this error tonight.
          Hide
          Marcus Clyne added a comment -

          Changing to these lines worked for me too:

          evfd = eventfd(0, 0);
          /* if (evfd < 0)
          ink_release_assert((evfd = eventfd(0,0)) >= 0);*/

          Show
          Marcus Clyne added a comment - Changing to these lines worked for me too: evfd = eventfd(0, 0); /* if (evfd < 0) ink_release_assert((evfd = eventfd(0,0)) >= 0);*/
          Hide
          Moshe Kaplan added a comment -

          Commenting these lines really solved this issue.
          Can you explain the HAVE_EVENTFD? Maybe another configuration should be chosen?

          #ifdef HAVE_EVENTFD
          evfd = eventfd(0, 0);
          /* if (evfd < 0)
          ink_release_assert((evfd = eventfd(0,0)) >= 0);*/
          fcntl(evfd, F_SETFD, FD_CLOEXEC);
          fcntl(evfd, F_SETFL, O_NONBLOCK);
          #else
          ink_release_assert(pipe(evpipe) >= 0);
          fcntl(evpipe[0], F_SETFD, FD_CLOEXEC);
          fcntl(evpipe[0], F_SETFL, O_NONBLOCK);
          fcntl(evpipe[1], F_SETFD, FD_CLOEXEC);
          fcntl(evpipe[1], F_SETFL, O_NONBLOCK);
          #endif

          Show
          Moshe Kaplan added a comment - Commenting these lines really solved this issue. Can you explain the HAVE_EVENTFD? Maybe another configuration should be chosen? #ifdef HAVE_EVENTFD evfd = eventfd(0, 0); /* if (evfd < 0) ink_release_assert((evfd = eventfd(0,0)) >= 0);*/ fcntl(evfd, F_SETFD, FD_CLOEXEC); fcntl(evfd, F_SETFL, O_NONBLOCK); #else ink_release_assert(pipe(evpipe) >= 0); fcntl(evpipe [0] , F_SETFD, FD_CLOEXEC); fcntl(evpipe [0] , F_SETFL, O_NONBLOCK); fcntl(evpipe [1] , F_SETFD, FD_CLOEXEC); fcntl(evpipe [1] , F_SETFL, O_NONBLOCK); #endif
          Hide
          Moshe Kaplan added a comment -

          In order to clarify I used the following Amazon AWS AMI:
          Basic Fedora Core 8 (AMI Id: ami-84db39ed)
          Minimal Fedora Core 8, 32-bit architecture, and Amazon EC2 AMI Tools.

          Show
          Moshe Kaplan added a comment - In order to clarify I used the following Amazon AWS AMI: Basic Fedora Core 8 (AMI Id: ami-84db39ed) Minimal Fedora Core 8, 32-bit architecture, and Amazon EC2 AMI Tools.
          Hide
          Jason Giedymin added a comment -

          For the record I can reproduce this error on EC2 using Fedora Core 8 kernel 2.6.21.7-2.ec2.v1.2.fc8xen.

          I was able to use George's code comment above and ripped out the assert for a quick test. ATS was then working correctly after I did this.

          Show
          Jason Giedymin added a comment - For the record I can reproduce this error on EC2 using Fedora Core 8 kernel 2.6.21.7-2.ec2.v1.2.fc8xen. I was able to use George's code comment above and ripped out the assert for a quick test. ATS was then working correctly after I did this.
          Hide
          George Paul added a comment -

          I should have mentioned to the change the assert check into printing out the 'errno' instead in case 'evfd < 0' so we can see why the syscall is failing.
          -G

          Show
          George Paul added a comment - I should have mentioned to the change the assert check into printing out the 'errno' instead in case 'evfd < 0' so we can see why the syscall is failing. -G
          Hide
          Moshe Kaplan added a comment -

          After updating the right files under /usr/local/etc/trafficserver/ the daemon started listening on port 80. However, I still did not had response from the mapped server.

          When I tailed /var/log/messages, I receive the following error that is mentioned in http://mail-archives.apache.org/mod_mbox/incubator-trafficserver-dev/201003.mbox/%3C1537511414.394401269193047225.JavaMail.jira@brutus.apache.org%3E (I really use 2.6.21.7-2.ec2.v1.2.fc8xen)

          Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_server[31067]:

          {1082005552} NOTE: cache clustering disabled
          Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_server[31067]: {1082005552}

          NOTE: clearing statistics
          Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_server[31067]: FATAL: UnixEThread.cc:70: failed assert `(evfd = eventfd(0,0)) >= 0`
          Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_manager[27640]:

          {3086059840} ERROR: [LocalManager::pollMgmtProcessServer] Server Process terminated due to Sig 6: Aborted
          Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_manager[27640]: {3086059840}

          ERROR: (last system error 2: No such file or directory)
          Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_manager[27640]:

          {3086059840} ERROR: [Alarms::signalAlarm] Server Process was reset
          Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_manager[27640]: {3086059840}

          ERROR: (last system error 2: No such file or directory)
          Mar 22 18:30:42 domU-12-31-39-09-25-24 traffic_manager[27640]:

          {3086059840} NOTE: [LocalManager::startProxy] Launching ts process
          Mar 22 18:30:42 domU-12-31-39-09-25-24 traffic_manager[27640]: {3086059840}

          NOTE: [LocalManager::pollMgmtProcessServer] New process connecting fd '9'
          Mar 22 18:30:42 domU-12-31-39-09-25-24 traffic_manager[27640]:

          {3086059840}

          NOTE: [Alarms::signalAlarm] Server Process born

          Please notice, that after I changed line 67 to the following code and remaked the code
          /* evfd = eventfd(0, O_NONBLOCK | FD_CLOEXEC);*/
          evfd = eventfd(0, 0);

          I received the error:
          FATAL: UnixEThread.cc:71: failed assert `(evfd = eventfd(0,0)) >= 0`
          (The error moved from line 70 to 71 due to the commented line, which means that server really runs the updated code)

          P.S I'm using the Amazon EC2 default EBS based Fedora core 8. Let me know if you need more information

          Best,
          Moshe Kaplan

          Show
          Moshe Kaplan added a comment - After updating the right files under /usr/local/etc/trafficserver/ the daemon started listening on port 80. However, I still did not had response from the mapped server. When I tailed /var/log/messages, I receive the following error that is mentioned in http://mail-archives.apache.org/mod_mbox/incubator-trafficserver-dev/201003.mbox/%3C1537511414.394401269193047225.JavaMail.jira@brutus.apache.org%3E (I really use 2.6.21.7-2.ec2.v1.2.fc8xen) Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_server [31067] : {1082005552} NOTE: cache clustering disabled Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_server [31067] : {1082005552} NOTE: clearing statistics Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_server [31067] : FATAL: UnixEThread.cc:70: failed assert `(evfd = eventfd(0,0)) >= 0` Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_manager [27640] : {3086059840} ERROR: [LocalManager::pollMgmtProcessServer] Server Process terminated due to Sig 6: Aborted Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_manager [27640] : {3086059840} ERROR: (last system error 2: No such file or directory) Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_manager [27640] : {3086059840} ERROR: [Alarms::signalAlarm] Server Process was reset Mar 22 18:30:41 domU-12-31-39-09-25-24 traffic_manager [27640] : {3086059840} ERROR: (last system error 2: No such file or directory) Mar 22 18:30:42 domU-12-31-39-09-25-24 traffic_manager [27640] : {3086059840} NOTE: [LocalManager::startProxy] Launching ts process Mar 22 18:30:42 domU-12-31-39-09-25-24 traffic_manager [27640] : {3086059840} NOTE: [LocalManager::pollMgmtProcessServer] New process connecting fd '9' Mar 22 18:30:42 domU-12-31-39-09-25-24 traffic_manager [27640] : {3086059840} NOTE: [Alarms::signalAlarm] Server Process born Please notice, that after I changed line 67 to the following code and remaked the code /* evfd = eventfd(0, O_NONBLOCK | FD_CLOEXEC);*/ evfd = eventfd(0, 0); I received the error: FATAL: UnixEThread.cc:71: failed assert `(evfd = eventfd(0,0)) >= 0` (The error moved from line 70 to 71 due to the commented line, which means that server really runs the updated code) P.S I'm using the Amazon EC2 default EBS based Fedora core 8. Let me know if you need more information Best, Moshe Kaplan
          Hide
          George Paul added a comment -

          Hi Marcus,

          Looking at the code path the problem is most likely a kernel version issue. From the eventfd() man page it looks like for linux kernels <=2.6.26 the 'flags' argument is unused and must be specified as zero. Try changing the line 'UnixEThread.cc:67' to

          evfd = eventfd(0, 0);

          recompile and test it out.

          regards,

          -George

          Show
          George Paul added a comment - Hi Marcus, Looking at the code path the problem is most likely a kernel version issue. From the eventfd() man page it looks like for linux kernels <=2.6.26 the 'flags' argument is unused and must be specified as zero. Try changing the line 'UnixEThread.cc:67' to evfd = eventfd(0, 0); recompile and test it out. regards, -George
          Hide
          Marcus Clyne added a comment -

          I've checked the same procedure with 2.0.1, and there are no problems on both my 64bit laptop and 32bit Virtuozzo VPS - only 32-bit 2.1.0 (not sure about other ones in the 2.1-branch).

          Show
          Marcus Clyne added a comment - I've checked the same procedure with 2.0.1, and there are no problems on both my 64bit laptop and 32bit Virtuozzo VPS - only 32-bit 2.1.0 (not sure about other ones in the 2.1-branch).
          Hide
          Marcus Clyne added a comment - - edited

          records.config
          ===========

          The records.config file is the only one that has been changed from the default one, and only two lines are different:

          CONFIG proxy.config.cluster.ethernet_interface STRING venet0:0

          • this is the virtualized network interface for the Virtuozzo platform and

          CONFIG proxy.config.url_remap.remap_required INT 0

          • to allow to work as a forward proxy (the only mode I'm trying to use it in)

          I have not attached any other files, since they are all untouched.

          install path
          ========

          The install path I've used is a custom one (/simpl/apps/trafficserver/trunk, which is linked to from /simpl/apps/trafficserver/default), and uses some symlinks - though it's working on my laptop using exactly the same setup script I've written.

          I will try installing in the default path to see if that makes a difference.

          Show
          Marcus Clyne added a comment - - edited records.config =========== The records.config file is the only one that has been changed from the default one, and only two lines are different: CONFIG proxy.config.cluster.ethernet_interface STRING venet0:0 this is the virtualized network interface for the Virtuozzo platform and CONFIG proxy.config.url_remap.remap_required INT 0 to allow to work as a forward proxy (the only mode I'm trying to use it in) I have not attached any other files, since they are all untouched. install path ======== The install path I've used is a custom one (/simpl/apps/trafficserver/trunk, which is linked to from /simpl/apps/trafficserver/default), and uses some symlinks - though it's working on my laptop using exactly the same setup script I've written. I will try installing in the default path to see if that makes a difference.
          Hide
          Jason Giedymin added a comment -

          Can you also post your configure settings? In particular your install path but attach everything.

          Show
          Jason Giedymin added a comment - Can you also post your configure settings? In particular your install path but attach everything.

            People

            • Assignee:
              George Paul
              Reporter:
              Marcus Clyne
            • Votes:
              0 Vote for this issue
              Watchers:
              0 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved:

                Development