[SOLVED] Odd things after V6 upgrade

[paped]

Hi - I am not sure if I have an issue here or not i.e. is it design intent in V6.0.1 so I am after some other users experiences and/or ideas for further testing/check of my system.

Background - I have a small home based system based on an Atom 1.6Ghz ITX system with 2Gb of ram and 1TB of HD storage - this box is also used for VPN, Samba storage. I have been using Zimbra (open source) since V4.x on this and an older 1Ghz spec ITX box in the past with no issues for a few years now. I have 5 accounts configured and have upgraded from V5.0.18 to 6.0.1 last night but I am having a few odd things happen.

Obviously with this being an Atom processor it is not that fastest box in the world but with previous Zimbra versions it takes a couple of seconds to login and maybe the odd second or so delay when open email but it is useable. When looking at HTOP there used to be maybe 5 to 10 processes related to Zimbra, the cpu was normally around 60% with odd peaks to 100% and the CPU load shown was around 1.5 to 1.75.

However after the upgrade to first thing I noticed was how slow the login was after entering the user name and password it takes between 15 and 40 seconds for the actual mailbox to come up so I started doing some basic (as not a linux guru) checking. When doing HTOP again I now see around 20 to 30 Zimbra services (at least) and there are 4 services which appear to be java based that seem to be requesting 100% of CPU most of the time to the point that the CPU is now stuck at 100% with only the occasional drop to 70-80%. The CPU load shown now seems to be around 3.75 to 4.0 constantly....

If this is design intent then fine I will live with it but if this is the case V6 does seem to have increased it's CPU footprint and number of services it runs by a number of magitudes, in fact to such a point that it has prompted this post as I am not sure if this is a problem or not....

The upgrade itself went well I had to add a line in to my host file and got one error around a port conflict on Zimbra store using port 443 which I use for another app but it did install ok, started ok and still seems to be working ok other then the slowness that appears to be related to the CPU being at 100%.

This upgrade though did install "memcache" as an additional app above what I had in 5.0.18? Not sure if this has anything to do with it but I have checked the log files and most of them look fine. The exception is the nginx.log where I have the following being add multiple times (from the time stamps on the entries this whole list of extries below is being repeated at least 30 times PER SECOND - I am sure that this may be using some of the CPU!!!!)

2009/09/16 06:55:42 [alert] 22559#0: socket() failed (97: Address family not supported by protocol)
2009/09/16 06:55:42 [error] 22559#0: cannot connect to memcached server 127.0.1.1:11211 (rc:-1)
2009/09/16 06:55:42 [notice] 6254#0: start worker process 22559
2009/09/16 06:55:42 [alert] 22560#0: socket() failed (97: Address family not supported by protocol)
2009/09/16 06:55:42 [error] 22560#0: cannot connect to memcached server 127.0.1.1:11211 (rc:-1)
2009/09/16 06:55:42 [notice] 6254#0: start worker process 22560
2009/09/16 06:55:42 [alert] 22561#0: socket() failed (97: Address family not supported by protocol)
2009/09/16 06:55:42 [error] 22561#0: cannot connect to memcached server 127.0.1.1:11211 (rc:-1)
2009/09/16 06:55:42 [notice] 6254#0: start worker process 22561
2009/09/16 06:55:42 [alert] 22562#0: socket() failed (97: Address family not supported by protocol)
2009/09/16 06:55:42 [error] 22562#0: cannot connect to memcached server 127.0.1.1:11211 (rc:-1)
2009/09/16 06:55:42 [notice] 6254#0: start worker process 22562
2009/09/16 06:55:42 [notice] 6254#0: signal 29 (SIGIO) received
2009/09/16 06:55:42 [notice] 6254#0: signal 17 (SIGCHLD) received
2009/09/16 06:55:42 [alert] 6254#0: worker process 22559 exited on signal 11
2009/09/16 06:55:42 [alert] 6254#0: worker process 22560 exited on signal 11
2009/09/16 06:55:42 [alert] 6254#0: worker process 22561 exited on signal 11
2009/09/16 06:55:42 [alert] 6254#0: worker process 22562 exited on signal 11

I am suspicious of the above because as noted above I seem to have 4 Zimbra processes trying to take 100% of CPU most of the time and there appears to be 4 processes in this log file entries restarting and failing constantly - is there anyway to turn Memcache off to stop these restarts/failures/log entries or do I need to do someelse to fix the problem and get it working properly?

So any observations or advice around this or additional testing that I can do would be appreciated.

Thanks,

Paul

[phoenix]

If this is a single server install then you don't need memcahced installed or running.

Check with the following:

zmprov gs `zmhostname` | grep zimbraServiceInstalled
zmprov gs `zmhostname` | grep zimbraServiceEnabled

If installed and running then get rid of it with:

zmprov ms `zmhostname` -zimbraServiceEnabled memcached
zmcontrol stop
zmcontrol start

[paped]

I first did the host file changes in this post which I found shortly after adding this thread - Zimbra nginx process defuncting 3000 times per second - this helped greatly as it stopped the log file entries and reduced some of the CPU usage.

Then I completed Phoenix's changes above to disable memcached and my system is now running like a charm. CPU running around 60 to 70% with a few 100% peak as before and average load about 1.50 to 1.60.

Thank you Phoenix and the poster in the above older link....