/ fs is full on controller (big /var/log/mongod.log)

Bug #1392593 reported by Sergey Galkin
This bug report is a duplicate of:  Bug #1393400: reduce logging verbosity of mongodb. Edit Remove
8
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Fuel for OpenStack
Confirmed
High
Matthew Mosesohn

Bug Description

astute_sha: 3c374c9f7bfbdbcd7ce2f716cd704e3044e6fb41
auth_required: true
build_id: 2014-11-12_22-00-23
build_number: '99'
feature_groups:
- mirantis
fuellib_sha: bc3c6298f203af3c911a6c5aa11760331c9bdd60
fuelmain_sha: 144262798c38c9186ad2d9c71c60a795b1474014
nailgun_sha: 563749b01dd26e1e0978636efa3a0dafa5745fab
ostf_sha: 720cc1308c3a7081736edd167e7928ca61914aaa
production: docker

Steps to reproduce:
1. Install cluster on 100 nodes (3 controllers in HA, 97 computes + ceph) wih ceilometer, ceph, neutron gre
2. Run tempest and rally tests

after ~8-9 hours uptime login to main controller, fs on / is full

[root@node-64 ~]# df -h
Filesystem Size Used Avail Use% Mounted on
/dev/mapper/os-root 15G 15G 0 100% /
tmpfs 16G 37M 16G 1% /dev/shm
/dev/md0 194M 27M 158M 15% /boot
/dev/mapper/mongo-mongodb
                      959G 54G 857G 6% /var/lib/mongo

[root@node-64 ~]# ls -lh /var/log/ | grep mongod.log
-rw------- 1 root adm 3.9G Nov 14 03:01 mongod.log
-rw------- 1 root adm 3.7G Nov 14 03:01 mongod.log.1
-rw------- 1 root adm 13M Nov 14 02:01 mongod.log.4.gz
-rw------- 1 root adm 19M Nov 14 01:01 mongod.log.5.gz

[root@node-64 ~]# uptime
 05:53:05 up 8:50, 0 users, load average: 2.54, 2.46, 2.61

Sorry, can't attache snapshot
[root@fuel rally]# du -sh /var/log/
9.7G /var/log/

Tags: scale
Sergey Galkin (sgalkin)
tags: added: scale
affects: mos → fuel
Revision history for this message
Tomasz 'Zen' Napierala (tzn) wrote :

Matt recently merged this: https://review.openstack.org/#/c/133956/ . MongoDB log should be rotated hourly according to configuration. We will investigate this

Changed in fuel:
status: New → Confirmed
importance: Undecided → High
assignee: nobody → Matthew Mosesohn (raytrac3r)
Revision history for this message
Matthew Mosesohn (raytrac3r) wrote :

They are rotated hourly. Look at the timestamps.

What's interesting is what level of logging is enabled on this system? Normal or debug? Also, rally generates far more traffic on a cluster than normal usage, so it may not affect most users.

Changed in fuel:
milestone: none → 6.0
Revision history for this message
Sergey Galkin (sgalkin) wrote :

Yes, debug enabled.
May be OS deployment script should allocate more space for /dev/mapper/os-root if debug enabled ?

Revision history for this message
Sergey Galkin (sgalkin) wrote :
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.