Bug 1368718
Summary: | qpid dispatch router on capsule leaking memory at scale | ||||||||
---|---|---|---|---|---|---|---|---|---|
Product: | Red Hat Satellite | Reporter: | Pradeep Kumar Surisetty <psuriset> | ||||||
Component: | Performance | Assignee: | satellite6-bugs <satellite6-bugs> | ||||||
Status: | CLOSED ERRATA | QA Contact: | Pradeep Kumar Surisetty <psuriset> | ||||||
Severity: | urgent | Docs Contact: | |||||||
Priority: | urgent | ||||||||
Version: | 6.2.0 | CC: | alexandre.chanu, arcsharm, bkearney, cduryee, dcaplan, egolov, jentrena, jhutar, mcressma, mmccune, omaciel, pmoravec, psuriset, tross, zhunting | ||||||
Target Milestone: | Unspecified | Keywords: | Performance, PrioBumpPM | ||||||
Target Release: | Unused | ||||||||
Hardware: | Unspecified | ||||||||
OS: | Unspecified | ||||||||
Whiteboard: | scale_lab | ||||||||
Fixed In Version: | qpid-dispatch-0.4-21, qpid-cpp-0.34-25 | Doc Type: | If docs needed, set a value | ||||||
Doc Text: | Story Points: | --- | |||||||
Clone Of: | |||||||||
: | 1463801 (view as bug list) | Environment: | |||||||
Last Closed: | 2017-08-10 17:02:29 UTC | Type: | Bug | ||||||
Regression: | --- | Mount Type: | --- | ||||||
Documentation: | --- | CRM: | |||||||
Verified Versions: | Category: | --- | |||||||
oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |||||||
Cloudforms Team: | --- | Target Upstream Version: | |||||||
Embargoed: | |||||||||
Attachments: |
|
Created attachment 1192485 [details]
qpid dispatch router mem on smaller capsule
A user was able to induce a qpidd/qrouterd OOM by registering/unregistering 100x, but with gofer 2.5. The user and myself both tried this with later versions of gofer but did not see any issue. For example, I tried the following which seemed to pass without OOM or other issue: for i in `seq 1 10000`; do subscription-manager register --username admin --password changeme --environment Library && subscription-manager unregister && sleep 2; done I will retry with 5k clients. Also, the test in #12 was direct against a satellite without any capsules. The 5K test will be against a capsule. (In reply to Chris Duryee from comment #12) > A user was able to induce a qpidd/qrouterd OOM by registering/unregistering > 100x, but with gofer 2.5. Did that test include goferd (re)start? If so, then imho bz1367735 hit. re #15, I don't know if goferd was restarted, it may have been. If so it explains the behavior. Havent noticed this lately. But noticed this again today Root cause: when we have different versions of qpid-dispatch-router on satellite and capsules, memory of qpid-dispatch-router shoots up for ex in my case; satellite : qpid-dispatch-router-0.4-21.el7sat.x86_64 capsuels: qpid-dispatch-router-0.6.1-5.el7.x86_64 We need to make sure, we have same version. Then we dont see this issue KCS is updated for this. Thats good enough Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2017:2466 |
Created attachment 1192484 [details] qpid dispatch router mem on bigger capsule Description of problem: Have registered 5000 content hosts aginst 2 capsules 1) with recommended h/w ( 2 cpus, 8G mem) 2) With better config ( 8 cpus, 16G mem) After registration is done, qpid-disptach-router memory is leaking. Its gradually increasing Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. Register content hosts at scale against capsule 2. 3. Actual results: Mmeory leak . qpid-disptach-router keep growing Expected results: lesser qpid-disptach-router memory consumption Additional info: