Martin,
Thank you. You might be on to something. I was quoting from memory and I
was wrong on swap. Of the two nodes, both in my mind identical VM's, the
secondary node has 8GB of swap and a tiny bit used, but the primary, the
one that is crashing, has no swap configured. I have requested our systems
team add 8GB of swap to the primary.
Primary server:
total used free shared buff/cache
available
Mem: 8010840 4872660 420488 107484 2717692
2679336
Swap: 0 0 0
Secondary server:
total used free shared buff/cache
available
Mem: 8010972 1192296 1530500 23196 5288176
6449948
Swap: 8388604 4604 8384000
Not sure I understand why it would matter since in theory swap should not
be needed on a server with 8GB of ram with jvm limit set to 6GB though.
Any more insight on why, because I would really like to understand the
reason.
Additionally, I've put the shibboleth IDP back into play, effectively
rendering the saml services in cas "unused". I am using proxy_ajp to front
tomcat with apache so it was easy to copy the idp.war into tomcat and
re-enable the shib-cas-authenticator. I guess my hope of moving from
cas+shibb. to just cas will have to wait....
Thanks,
William
P.S. Jeff, thank you for posting your catalina opts!
On Wednesday, December 20, 2017 at 11:30:40 PM UTC-6, Martin Bohun wrote:
>
> I have seen the behavior you are describing when people ran cas (tomcat,
> mysql, etc.) on a (what I would consider a misconfigured) Linux box with 0
> swap.
> However you are saying you have 4gb of swap.
> I still do prefer to set my swap to 2 * $MY_RAM; can you try that? adjust
> or add a swapfile to your swap (so you have 8gb RAM / 16gb swap), I am
> curious if that would help / solve your problem?
> What error messages are you getting in the jvm and syslog/systemd journal
> from the OS?
>
> regards,
>
> martin
>
> On Thursday, December 21, 2017 at 1:35:45 PM UTC+11, William E. wrote:
>>
>> RHEL 7, 8GB ram, swap is 4GB. It's a VM in our vSphere cluster+SAN. I
>> actually have three, two PROD nodes behind a load balancer and one test
>> node. All have same specs and all show the issue. Steadily chews up
>> memory until eventual crash, 1-6 hours depending on load.
>>
>> The asme servers were running cas 3.6 . + shibboleth 3.3.x for quite a
>> while without memory issues. Upgraded and tried to consolidate to just cas
>> 5, using it's saml2 capabilities to replace the shibboleth component. But,
>> it's not going as well as I had hoped.
>>
>> Been working with Unicon Support on it, but it appears to be a memory
>> leak in cas 5.2, based on heap analysis. So I am kinda of stuck.
>>
>> Thanks for your help!
>>
>>
>>
>> On Wednesday, December 20, 2017 at 6:49:39 PM UTC-6, Martin Bohun wrote:
>>>
>>> What is your:
>>> 1. operation system
>>> 2. how much RAM do you have
>>> 3. how much swap do you have
>>>
>>> if you are on Linux you can do:
>>> 1. uname -a
>>> 2-3. free -m
>>>
>>> and post the output here
>>>
>>> regards,
>>>
>>> martin
>>>
>>> On Thursday, December 21, 2017 at 11:00:30 AM UTC+11, William E. wrote:
>>>>
>>>> Does anyone have any recommendations for CATALINA_OPTS for cas 5.x on
>>>> tomcat 8?
>>>>
>>>> I am finding that our setup steadily eats up memory to the point that
>>>> it eventually crashes from out of memory and has to be restarted.
>>>>
>>>> Current settings:
>>>>
>>>> CATALINA_OPTS="-Djava.awt.headless=true -Dfile.encoding=UTF-8 -server
>>>> -Xms1g -Xmx6g -XX:-UseGCOverheadLimit -XX:+UseConcMarkSweepGC
>>>> -XX:-UseCompressedOops"
>>>>
>>>> JAVA_OPTS=$CATALINA_OPTS
>>>>
>>>>
>>>> Thanks,
>>>> William
>>>>
>>>>
--
- Website: https://apereo.github.io/cas
- Gitter Chatroom: https://gitter.im/apereo/cas
- List Guidelines: https://goo.gl/1VRrw7
- Contributions: https://goo.gl/mh7qDG
---
You received this message because you are subscribed to the Google Groups "CAS
Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/a/apereo.org/d/msgid/cas-user/7e36f7d2-3bf7-49d2-bcd8-bbc0e22b901b%40apereo.org.
