Hi, after some days i noticed that our monitoring host lost memory over time. When investigating it will only be cured by reboot so we are most likely in kernel space here looking for a bug.
Investigating further my guess is that the dentry slab grows over time. From my
calculations my idea is that check_by_ssh looses a dentrys per invocation.
Does anyone else see this behaviour?
root@monitoring:~# egrep "num_objs|dentry" /proc/slabinfo
# name <active_objs> <num_objs> <objsize> <objperslab>
<pagesperslab> : tunables <limit> <batchcount> <sharedfactor> : slabdata
<active_slabs> <num_slabs> <sharedavail>
dentry 13714110 13714575 192 21 1 : tunables 0 0 0
: slabdata 653075 653075 0
Uptime is ~9 days - ~1800 checks at 5 Minutes.
flo@p3:~$ echo $[ 9 * 24 * 60 * 1800 / 5 ]
4665600
The machine has no remote filesystems and has a total of ~500K files so we
cant fill dentrys with files on disk.
root@monitoring:~# uname -a
Linux monitoring.filoo.de 3.13.0-122-generic #171-Ubuntu SMP Tue Jun 20
16:43:52 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux
Dropping the kernel caches solves the Problem:
sync; echo 3 > /proc/sys/vm/drop_caches
Flo
--
Florian Lohoff f...@zz.de
UTF-8 Test: The 🐈 ran after a 🐁, but the 🐁 ran away
signature.asc
Description: Digital signature
_______________________________________________ icinga-users mailing list icinga-users@lists.icinga.org https://lists.icinga.org/mailman/listinfo/icinga-users
