max_user_instances in inotify runs low on juju-db units - version 1.25.6
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
juju-core |
New
|
Undecided
|
Unassigned |
Bug Description
I found on one of our clouds that is back-rev at 1.25.6 that there is an issue with running many services and running out of inotify user instances.
Issue was found when running tail -f /var/log/
tail: inotify cannot be used, reverting to polling: Too many open files
Tried upping max_user_watches from 1024 to 1048576 (per suggestion on https:/
Didn't help at that point, still tail -f showed the too many open files.
I upped the fs.inotify.
I think juju-db may need to come with a raised max_user_instances setting to account for large cloud installs.
running ubuntu trusty with all updates (other than juju agent versions)
List of processes and filehandles pointing to inotify:
root@lexisnexis
8 /proc/8567/
8 /proc/8565/
8 /proc/7829/
8 /proc/6676/
8 /proc/6606/
8 /proc/1356295/
8 /proc/1182922/
7 /proc/8560/
6 /proc/8376/
6 /proc/7601/
6 /proc/2676260/
6 /proc/1/
6 /proc/1787829/
2 /proc/7763/
2 /proc/6737/
1 /proc/5996/
1 /proc/4348/
1 /proc/4146642/
1 /proc/4055301/
1 /proc/4038967/
1 /proc/37736/
1 /proc/36633/
1 /proc/36592/
1 /proc/35544/
1 /proc/34577/
1 /proc/33171/
1 /proc/32643/
1 /proc/31746/
1 /proc/31610/
1 /proc/30852/
1 /proc/29809/
1 /proc/2678148/
1 /proc/2251006/
1 /proc/2189625/
1 /proc/2187232/
1 /proc/2150373/
1 /proc/2053725/
1 /proc/2052087/
1 /proc/2013083/
1 /proc/2008976/
1 /proc/1999224/
1 /proc/1985370/
1 /proc/1982901/
1 /proc/1789709/
1 /proc/1359607/
1 /proc/1184707/
FYI smells like lp#1631038 , which looks like fixed juju 2.0+