Too many logins from ClusterController into cluster nodes
Recently I am facing too many logins on the nodes coming from the clustercontroller .
As a result I am facing a high cpu usage due to the accounts-daemon process.
This is becomming annoying.
Any idea how clustercontroler gets his statistics data from the nodes?
Is this issue related to the polling of these info?
-
Official comment
Hi,
Can you send a
"SHOW PROCESSLIST"
orselect * from information_schema.processlist where user='cmon';
From one of the affected db nodes?Also what cmon version are you using?
BR
johanComment actions -
on an affected node:
3250, 'cmon', '192.168.255.143:54914', 'information_schema', 'Sleep', 0, '', '', 115.361, 0, 0, 0.000, 67408, 0, 117062, '', 26803
3249, 'cmon', '192.168.255.143:54864', 'information_schema', 'Sleep', 2, '', '', 2443.765, 0, 0, 0.000, 67408, 0, 117040, '', 29626
cmon version:
cmon, version 1.3.2.1391
Thank You
Alan
-
Hi,
Also please checkout this one (re: accounts-daemon):
http://severalnines.com/blog/clustercontrol-tips-tricks-wtmp-log-rotation-settings-sudo-user
A large wtmp file will cause the accounts daemon to act up b/c the accounts daemon is buggy and will try and re-read the wtmp fully as soon as a login / logout happens. If this then is a big file then the accounts daemon will eat cpu.
This in turn could lead to connection issues.
BR
johan
Please sign in to leave a comment.
Comments
7 comments