Good day all
Hoping this mail finds you well.
I am hoping someone can perhaps give us some guidance here as we now seem to
be stuck on a problem and have not been able to find a solution after more than
a month.
We are running an opensips server on Centos 6.5 , using mysql 5.7.13 which
was installed via Tarball file.
The server is setup as a slave and master and receives updates from opensips
config nodes as well as registrations from workers.
Replication is paused during the day and forward replication (master) is
disabled at the moment.
However , we are getting an issue every day on mysql side in terms of mysql
pushing up server load.
During the day the server is running fine with a load avg not going above
1.5 during peak times.
However in the evening , replication is unpaused, and completes processing
and catchup within about 15 minutes and is paused again about 30 minutes after
the unpause.
Give or take 45 minutes to an hour after the replication is paused again,
mysql starts to cause high cpu usage with no apparent processes running as can
be seen on full processlist (maybe one or two selects which completes fairly
quickly)
The higher load, causes queries to slow down however and opensips to
start timing out on db connections, causing clients to resubmit.
The resubmits , then obviously causes even more load spiking the mysql
load to increase as well as the server load and eventually opensips kills
itself.
I have looked at the disks, iowaits, memory usage, all is fine.
We do not see any strange queries or stick queries, no deadlocks, etc...
only the increase in selects after mysql starts to push up the cpu load.
We have added all indexes we can find, but even this has made no difference
at all.
Currently we are at a loss so I am hoping someone else can assist in
explaining how else we can find out why mysql is eating up the cpu ...
The same behaviour can also be seen the moment any new feature is added to
the server that requires mysql processing to be done, so this does not seem to
be specifically related to replication, however it does seem like the current
load from replication causes mysql to act up.
the server is currently running on SSD (recently replaced) , and 8Gb of
memory with 1 x quadcore CPU.
should any more info be required, please feel free to ask.
