8.7.11 -> 8.8.5 -> 8.8.6 (OSE CentOS7) problem

[mkuchin]

I upgraded from OSE 8.7.11 to 8.8.5
passed without errors
After the update, I noticed a periodically occurring problem
Some time the services work, then they work with delays, then they stop responding
this applies to both the web and the imap
After that I saw that 8.8.6 is available and updated to this version
The problem began to manifest itself later
The services are working, after 4 hours delays begin, after 6 hours the services are not responding.

HTTP ERROR 502

Problem accessing ZCS upstream server. Cannot connect to the ZCS upstream server. Connection is refused.
Possible reasons:

upstream server is unreachable
upstream server is currently being upgraded
upstream server is down

Please contact your ZCS administrator to fix the problem.

Powered by Nginx-Zimbra://

restarting the service fixes the problem again for a few hours
system - centos 7, updates are made regularly
resources is enough, 8.7.11 worked perfectly
In the logs nothing particularly critical was noticed
Where can be the problem?

Thanks
Maxim

[mkuchin]

I saw that at the time of the problem, Java uses 102% cpu
20 mailboxes

zmprov ms `zmhostname` zimbraImapNumThreads 50
zmprov ms `zmhostname` zimbraHttpNumThreads 60
zmprov mcf zimbraLmtpNumThreads 20

these settings did not help

[mkuchin]

zmprov ms `zmhostname` zimbraImapNumThreads 20
zmprov ms `zmhostname` zimbraHttpNumThreads 20
zmprov mcf zimbraLmtpNumThreads 20

This one makes zimbra working normally in my case

[comradeTJH]

Quite a few users including me are experience a similar problem. Nothing seems to fix the problem. It's also extremely hard to debug since just the java/jetty processes kind of get stuck. Everything stops to respond and only a 'zmmailboxdctl restart' fixes the problem temporarily.

[mkuchin]

zmprov ms `zmhostname` zimbraImapNumThreads 20
zmprov ms `zmhostname` zimbraHttpNumThreads 20
zmprov mcf zimbraLmtpNumThreads 20

This one makes zimbra working normally in my case

all the same the problem is not solved. it just took longer to work. restart the service is still necessary

[msquadrat]

You should try the hotfix which was published yesterday: https://wiki.zimbra.com/wiki/Zimbra_Rel ... 8.6_HotFix

[comradeTJH]

Ah yes, thanks! This seems to do the trick. No more sleepless nights for the moment

[mkuchin]

Well, it seems working
Thanks!

[mkuchin]

when sending a message, an error occurs

method: [unknown]
msg: try again: Unable to connect to the MTA
code: mail.TRY_AGAIN
detail: soap:Receiver
trace: qtp998351292-3795:1518864511437:f26455ff63936be7