Too many open files error in Jira server
Symptoms
A JIRA application experiences a general loss of functionality in several areas.
The following error appears in the atlassian-jira.log:
java.io.IOException: java.io.IOException: Too many open files
at java.lang.UNIXProcess.<init>(UNIXProcess.java:148)
at java.lang.ProcessImpl.start(ProcessImpl.java:65)
at java.lang.ProcessBuilder.start(ProcessBuilder.java:451)
at java.lang.Runtime.exec(Runtime.java:591)
at java.lang.Runtime.exec(Runtime.java:429)
at java.lang.Runtime.exec(Runtime.java:326)
at org.netbeans.lib.cvsclient.connection.LocalConnection.openConnection(LocalConnection.java:57)
at org.netbeans.lib.cvsclient.connection.LocalConnection.open(LocalConnection.java:110)
at com.atlassian.jira.vcs.cvsimpl.CvsRepositoryUtilImpl.openConnectionToRepository(CvsRepositoryUtilImpl.java:443)
at com.atlassian.jira.vcs.cvsimpl.CvsRepositoryUtilImpl.updateCvs(CvsRepositoryUtilImpl.java:244)
at com.atlassian.jira.vcs.cvsimpl.CvsRepository.updateCvs(CvsRepository.java:241)
at com.atlassian.jira.vcs.cvsimpl.CvsRepository.updateRepository(CvsRepository.java:298)
at com.atlassian.jira.vcs.DefaultRepositoryManager.updateRepository(DefaultRepositoryManager.java:657)
at com.atlassian.jira.vcs.DefaultRepositoryManager.updateRepositories(DefaultRepositoryManager.java:608)
at com.atlassian.jira.service.services.vcs.VcsService.run(VcsService.java:55)
at com.atlassian.jira.service.JiraServiceContainerImpl.run(JiraServiceContainerImpl.java:67)
at com.atlassian.jira.service.ServiceRunner.execute(ServiceRunner.java:61)
at org.quartz.core.JobRunShell.run(JobRunShell.java:191)
at org.quartz.simpl.SimpleThreadPool$WorkerThread.run(SimpleThreadPool.java:516)To identify the current open file handler count limit please run:
ulimit -aS | grep openAnd then look at the row which looks like:
open files (-n) 2560 Cause
Lucene, the indexing system that is used by JIRA applications, does not support NFS mounts in JIRA applications. Using a NFS mount is known to cause this behaviour - further information can be found on the Lucene documentation.
UNIX systems have a limit on the number of files that can be concurrently open by any one process. The default for most distributions is only 1024 files, and for certain configurations of JIRA applications, this is too small a number. When that limit is hit, the above exception is generated and JIRA applications can fail to function as it cannot open the required files to complete its current operation.
This usually happens in JIRA application instances with particularly large heap space memory allocations. On each search, a lock file will be placed on the file system and deleted, however, the handle will not be released by JAVA until the garbage collector runs a Full Garbage Collection, removes these dereferenced objects, and clears the file handle through each object's finalize() method. If there has not been a collection for some time, it can precipitate this error.
Additionally, the following bugs are reported that are known to cause this behaviour:
- JRA-29587 - Getting issue details... STATUS
- JRA-35726 - Getting issue details... STATUS
- JRA-39114 - Getting issue details... STATUS (affects JIRA 6.2 and higher)
We have the following improvement requests to better handle this in JIRA applications:
Resolution
If you are using a JIRA application version with the bug in JRA-29587 - Getting issue details... STATUS , upgrade it to the latest version. If using NFS, migrate to a local storage mount.
For a more permanent solution of increasing the number of open files, see your operating system's manual. For example, Ubuntu users would follow the below.
- Modify the limits.conf file with the following command:
sudo vim /etc/security/limits.conf Add the following for the user that runs JIRA applications. If you have used the bundled installer, this will be
jira.limits.conf#<domain> <type> <item> <value> # #* soft core 0 #root hard core 100000 #* hard rss 10000 #@student hard nproc 20 #@faculty soft nproc 20 #@faculty hard nproc 50 #ftp hard nproc 0 #ftp - chroot /ftp #@student - maxlogins 4 jira soft nofile 16384 jira hard nofile 32768Modify the common-session file with the following:
sudo vim /etc/pam.d/common-session
common-sessionis a file only available in debian/ubuntuAdd the following line:
common-session# The following changes were made from the JIRA KB (https://confluence.atlassian.com/display/JIRAKB/Loss+of+Functionality+due+to+Too+Many+Open+Files+Error): session required pam_limits.so- Restart the JIRA server application.
For exceedingly large instances, we recommend consulting with our partners for scaling JIRA applications. See Is Clustering or Load Balancing JIRA Possible.
This changes will only work on installation that uses built in initd script for starting Jira. For installations that uses custom build service for systemd (latest versions of linux OS-es) changes will need to be applied directly in that systemd service configuration in a form of:
[Service]
LimitNOFILE=20000Diagnosis
In order to identify what files are kept open, the lsof +L1 command can be used. This will provide a list of the open files, for example:
COMMAND PID USER FD TYPE DEVICE SIZE/OFF NLINK NODE NAME
java 2565 dos 534r REG 8,17 11219 0 57809485 /home/dos/deploy/applinks-jira/temp/jar_cache3983695525155383469.tmp (deleted)
java 2565 dos 536r REG 8,17 29732 0 57809486 /home/dos/deploy/applinks-jira/temp/jar_cache5041452221772032513.tmp (deleted)
java 2565 dos 537r REG 8,17 197860 0 57809487 /home/dos/deploy/applinks-jira/temp/jar_cache6047396568660382237.tmp (deleted)Atlassian Support can investigate this further if the resolution does not work, in order to do so please provide the following to Atlassian Support.
- The file generated by
lsof +L1 > open_files.txt.
This command must be executed by the JIRA user, or a user who can view the files. For example if JIRA is running as root(which is not at all recommended), executing this command asjirawill not show the open files. - A heap dump taken at the time of the exception being thrown, as per Generating a Heap Dump.
- A JIRA application Support ZIP.