Red Hat Bugzilla – Bug 437017
Satellite suffers from severe performance degradation during System Group changes
Last modified: 2009-05-28 16:08:45 EDT
Description of problem:
Removing a host from a System Group via the RHN Satellite web interface
generates severely high loads and very high %iowait due to excessive disk
writes. Removal of a single host from a System Group takes 5 minutes. Removal of
N hosts takes 5N minutes.
Version-Release number of selected component (if applicable):
RHN release 4.1.5
Steps to Reproduce:
1. Remove host from System Group
5 minutes of high load and %iowait while system writes copious amounts of data
Low load, low disk usage, quick return time.
sar data is attached.
For reference, a single host was removed from the USWHWK6 System Group at
16:33:01. Time to completion: 5+ minutes. At 16:41:05, two hosts were removed
from the same System Group. Time to completion: 9+ minutes.
Note spike in loadavg, %iowait, block device IO during executions.
Created attachment 297647 [details]
sar data captured during removal of one, and then two, hosts from a single System Group
More info on this Satellite server:
1594 subscribed hosts - all with Management Entitlement. A couple (RHN Proxies)
193 System Groups
22 Software Channels, 10 of these are custom. the rest standard Red Hat channels.
Some of the custom channels are Clones. Some are not.
403 managed packages (uploaded via rhnpush).
No config channels defined.
Even more info on satellite server. I'm not sure if these numbers are high or
low or not:
[user@host ~]$ sudo find /data01/satellite/ -type f | wc -l
[user@host ~]$ sudo find /data01/satellite/ -type l | wc -l
[user@host ~]$ sudo find /data01/satellite/ -type d | wc -l
has there been a support request opened for this to track with?
Right now there is no customer associated ticket aligned to this bug report, nor the initial investigation performed by our Support Engineering Group to gain replication and understanding of the bug being reported.
Hey Cliff -
Not sure if there's been a case opened or not. Considering that the report is a year old at this point, we've upgraded to Sat 5.1, and I don't think we've seen this problem, it's fairly safe to close this BZ.