Server idle yet memcached.exe and erl.exe CPU use is extremly high

itay · September 12, 2015, 6:59am

Just saw your answer.

I’m a bit afraid of using RC. Is it stable and bug free ?
3.0.1 CE is really giving me a bad time so I want to make sure that I’m making the right decisions, as upgrading the DB is a delicate operation.

Will the full 4.0.0 version be released CE as well ? Is there a time frame ?
Will 3.0.3 be CE soon ?
Is there a page where you publish your versions road map ?

Thanks,

cihangirb · September 12, 2015, 10:04pm

Hi @itay, The timeline for 4.0 CE is early Oct. 4.0 includes all fixes in 3.0.3 version so I would recommend 4.0 CE when we are generally available. RC0 is not the finished version so we still have a few fixes we are adding before the targeted oct date.
thanks
-cihan

itay · September 15, 2015, 8:01am

Excellent
Does 4.0 requires any code modifications or is it plug and pray ?

cihangirb · September 15, 2015, 3:46pm

4.0 is fully backward compatible unless you are using an un-documented method or you don’t know about the items we have here in the deprecation list (below),
http://developer.couchbase.com/documentation/server/4.0/release-notes/deprecated.html

itay · September 16, 2015, 5:08am

Thanks,
I’m using .NET SDK so I guess there will be a specific version for 4.0 that will notify on any deprecated feature by not implementing them

itay · October 26, 2015, 6:01pm

@cihangirb,

Before migrating from 3.0.1 CE to 4.0.0 CE, where can I find a guide, manual, walkthru on how to do it correctly and successfully for the first time, on a running cluster ?

Thanks, Itay

cihangirb · October 26, 2015, 8:40pm

Hi @itay, pls rehearse on another cluster first to make sure things work well. Also review the known issues list to ensure there isn’t anything that would impact your specific case.

You can find the upgrade manual here.
There are 3 upgrade strategies listed on the page and all 3 strategies mentioned here will work with CE.
Thanks
-cihan

itay · October 27, 2015, 5:21am

Thanks @cihangirb,

@jmorris, Should I upgrade from .NET SDK 2.1.3 to 2.2.1 before or after I upgrade to 4.0.0 CE ?

jmorris · October 28, 2015, 12:18am

@itay -

I would move to 2.2.1 after you upgrade to 4.0.0 CE.

Thanks,

Jeff

itay · October 28, 2015, 5:36am

Thanks,
I’ll first upgrade the cluster and then the SDK

itay · October 28, 2015, 6:29am

@cihangirb,

I read “the manual” but found nothing as to what I should do to upgrade.

What should I do once I download the new version ? Perhaps the answer is trivial, as in “just double click it” and everything should be taken care of.
Is there a step-by-step guide ?

cihangirb · October 28, 2015, 6:29pm

you are right the instructions aren’t super clear here. http://developer.couchbase.com/documentation/server/4.0/install/upgrade-swap.html

Online Swap Rebalance is what I recommend you rehearse for your upgrade:

Take a new node, install 4.0 (dpkg -i …)
On your 2.x or 3.x cluster, remove an existing node and add the new node that has 4.0 on it.
Rebalance.
repeat until all nodes are running 4.0 binaries.

thanks
-cihan

anil · October 28, 2015, 6:58pm

Here are the documentation step-by-step guide for upgrade -

Thanks,

Anil Kumar

itay · October 28, 2015, 7:36pm

Thanks, I have some reading to do now

itay · October 29, 2015, 9:24am

@cihangirb,

I performed Single node upgrade and installation went very smoothly !
However, 1 hour later (!) cluster is still pending. To what, I cannot tell.

Update: 7 hours later still pending. Tried restart, still pending

Continuing here

Crapulax · December 23, 2015, 10:53am

Hello,

I was running into this issue on windows of high unexpected cpu usage of both erl and memchached. I upgraded to couchbase version 4.0.0-4051 Community Edition (build-4051) where I understand this issue is supposed to be corrected. And in my case it is not.
I am in dev environment and my couchbase server even when it does nothing run into the problem several times a day. Looking today in logs I might have some clue on what is happenning. It seems that the moxi server exits with a status 255 after having run 2h02mn (exactly!, idle socket timeout?):
Service ‘moxi’ exited with status 255. Restarting. Messages: e:\Bubble\CouchBase\bin\event_core.dll(event_base_loop+385) [0x00007FFA587DC571]
[0x00007FF64B748196]
[0x00007FF64B7843C7]
C:\WINDOWS\system32\KERNEL32.DLL(BaseThreadInitThunk+34) [0x00007FFA6E818102]
C:\WINDOWS\SYSTEM32\ntdll.dll(RtlUserThreadStart+52) [0x00007FFA70B4C2E4] ns_log000 ns_1@127.0.0.1 10:56:10 - Wed Dec 23, 2015

Bucket “bubble” loaded on node ‘ns_1@127.0.0.1’ in 0 seconds. ns_memcached000 ns_1@127.0.0.1 08:54:33 - Wed Dec 23, 2015

Is there a way to deactivate this server, we are using the sync gw and I understand and guess that in this case it goes through the direct sdk connection.
Note that this problem is making couchbase server unusable on Windows…As beside the CPU usage the sync gw is not anymore communicating with couchbase server correctly…

Thanks!

Crapulax · February 20, 2016, 1:31pm

Unlike what I wrote the moxi server was not the issue, killing it before the fatal 2h was not solving the issue.
Strange enough, I am not running into the issue anymore. What have changed since then is:

using Sync Gateway 1.2 instead of 1.1.1 or 1.1
filling up the couchbase server quite a lot
This is very weird.