Thursday, 2018-07-12

*** liyi_ has quit IRC00:37
*** liyi has joined #senlin00:37
*** fabian_ has joined #senlin00:55
*** blkart has quit IRC01:57
*** blkart has joined #senlin01:57
*** blkart has quit IRC02:01
*** blkart has joined #senlin02:10
*** fabian_ is now known as chenyb402:16
openstackgerritMerged openstack/senlin master: Fix stop node before delete error handling  https://review.openstack.org/58077302:32
openstackgerritMerged openstack/senlin master: Add cluster.stop_node_before_delete documentation  https://review.openstack.org/58052002:32
*** jmlowe has quit IRC02:35
chenyb4hi, dtruong02:58
*** dtruong2 has joined #senlin03:16
dtruong2Hi chenyb403:17
chenyb4dtruong2, About this patch https://review.openstack.org/#/c/521892/,  I have no idea start listener03:18
chenyb4dtruong2, Engine start, the listener will begin running03:21
chenyb4dtruong2, if event listener cluster enable health check,execute recover operation.03:22
*** dtruong2_ has joined #senlin03:24
*** liyi has quit IRC03:25
*** dtruong2 has quit IRC03:34
chenyb4dtruong2_, Can you give me advice?03:45
dtruong2_i was looking at the original bug report03:48
dtruong2_trying to understand what the problem was03:48
dtruong2_I think we should just follow what other projects are doing for notification endpoints:03:55
dtruong2_http://git.openstack.org/cgit/openstack/searchlight/tree/searchlight/listener.py#n17903:55
dtruong2_most projects start the the listener inside their start() function03:55
dtruong2_and stop and wait for the listener inside their stop() function03:56
chenyb4dtruong2_, Looks like this patch was fix it. https://review.openstack.org/#/c/521892/  Only the listener start or stop func name not same.04:02
dtruong2_so does it work with that patch?04:04
chenyb4Yes, that patch work ok.04:08
dtruong2_so what is the question then?04:09
chenyb4Ruiji wants to optimize the event health, start the listener when the engine starts, recover according to the listener's listener message, remove each engine will start a listener。04:13
dtruong2_it should be already doing that04:14
dtruong2_because it is running dummy_task as a timer04:14
dtruong2_so it loads the registry every so often and checks if the health manager needs to add a listener04:15
dtruong2_for example if there is engine A and B running, and engine A has the listener added04:17
dtruong2_if engine A dies, then the health manager in engine B will detect that and add the listener when it runs the dummy_task04:18
chenyb4So here only need to implement the call?04:24
dtruong2_To me what you have in the patch should be enough to fix the bug04:30
dtruong2_unless there is something that i'm missing04:31
dtruong2_but i have never used the health policy with lifecycle events so you will have test it and make sure it works ok04:32
dtruong2_when i tested health policy with node poll, i tried with two engines and tested what happens when one engine dies04:33
dtruong2_the other engine was able to pick up the health check and continue doing the node poll04:34
chenyb4That patch work ok and test is right.04:34
dtruong2_i see ruijie comment in the patch that listener.wait() blocks the current thread04:43
dtruong2_did you have any problems like that?04:43
chenyb4dtruong2_, that is he want to optimize the event health. So he commit https://review.openstack.org/#/c/529457/ split event health check.04:55
dtruong2_Interesting.  I never noticed the separate endpoint implementation.05:13
dtruong2_Does that implementation work?  It looks like it's not all finished05:18
dtruong2_Because from the looks of it, I don't think the plugin will solve the problem in the original bug report05:26
chenyb4Yes, ruijie commit patch change event health check not finish05:27
chenyb4he was commit 3 patch in event health check. https://review.openstack.org/#/c/529457/  https://review.openstack.org/#/c/529279/  https://review.openstack.org/#/c/529235/05:29
*** liyi has joined #senlin05:44
dtruong2_i'm sure how that event health check would make it more efficient, unless he has time to continue it.05:46
dtruong2_i can try out your patch tomorrow and see if the wait call is blocking the thread05:47
chenyb4Yes, I think so.05:48
chenyb4I think use that patch fix event health first.  https://review.openstack.org/#/c/521892/05:48
chenyb4thanks05:48
*** dtruong2_ has quit IRC05:57
openstackgerritchenyb4 proposed openstack/python-senlinclient master: Change CURRENT_API_VERSION to "1.9"  https://review.openstack.org/58196306:42
*** liyi has quit IRC07:20
*** liyi has joined #senlin07:24
openstackgerritDuc Nguyen proposed openstack/senlin-dashboard master: Add py36 testenv  https://review.openstack.org/58205708:05
*** chenyb4 has quit IRC10:25
*** liyi has quit IRC11:02
*** liyi has joined #senlin12:06
*** liyi has quit IRC12:11
*** liyi has joined #senlin13:03
*** liyi has quit IRC13:08
*** chenyb4 has joined #senlin13:52
*** chenyb4 has quit IRC14:24
*** chenyb4 has joined #senlin14:40
*** jmlowe has joined #senlin15:02
*** liyi has joined #senlin15:14
*** liyi has quit IRC15:19
*** jmlowe has quit IRC15:47
openstackgerritchenyb4 proposed openstack/python-senlinclient master: Change CURRENT_API_VERSION to "1.10"  https://review.openstack.org/58196316:29
openstackgerritchenyb4 proposed openstack/senlin master: Fix doc history.rst format error  https://review.openstack.org/58230617:14
*** chenyb4 has quit IRC17:24
openstackgerritDuc Truong proposed openstack/senlin master: Fix doc format errors  https://review.openstack.org/58231617:50
openstackgerritDuc Truong proposed openstack/senlin master: Treat sphinx warnings as errors  https://review.openstack.org/58232018:02
*** liyi has joined #senlin18:23
*** jmlowe has joined #senlin18:24
*** liyi has quit IRC18:27
*** liyi has joined #senlin20:32
*** liyi has quit IRC20:36
openstackgerritDuc Truong proposed openstack/senlin master: [DNM] Test sphinx CI job  https://review.openstack.org/58237920:53
*** liyi has joined #senlin22:22
*** liyi has quit IRC22:26

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!