-
Notifications
You must be signed in to change notification settings - Fork 335
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
fix #1597 #1623
base: master
Are you sure you want to change the base?
fix #1597 #1623
Conversation
Thanks, I'll give a look and ifwe got what we need for the regenerators tests so we can add a fail one and apply this fix :) |
otherwise very good catch&job already :) |
@gst ,i have fixed the problem you proposed, you can have a look at the code and if something is still wrong, please let me know. |
but first I'm gonna try to reproduce the problem on my side.. though I'm not completely sure that I've an exact/good view/understanding of it.. we'll see :) |
With this the Livestatus was simply not usable at all within a scheduler :| Seen while investigating on shinken-solutions/shinken#1623
damn.. The function modified here ( ho boy.. have to "untie" this kind of spaghettis situation.. again :s |
+1 for you @andyxning |
@gst , i have remove all the |
great @andyxning (that looks like already slightly better, isn't it ?, hold on there is still better, see bottom.. ;)) but as I explained in my previous comment : I'm rather unsure that the proposed changed will not fix the problem of memory consumption in/with the broker daemon simply because the code/function involved here is only called (as far as I see) if LiveStatus or WebUi are used in the scheduler, and not the broker.. Now maybe this would or could also fix some similar things in the scheduler case.. so this is tbc.. ;) Now back to this commit and this specific part within the function :
this is completely equivalent to :
which is again, now clearly (I hope you agree ;)), better than the previous already slightly better form.. ;) but again, I say that only/more for information/teach purpose because again : I'm unsure about all this unfortunately :( though if that would fix some things in your actual setup/configration regarding the memory problems we are hunting then I'd be happy to see that. I will be able to get back on this only by next week. |
thanks for the feedback. I will think about and refactor this again. To make more clear, i have not enable regenerator module for Scheduler, so However, in my opinion, we always fix the problem that the main reason to
as the comments said that we have fix the notificationways double problem All the other works are used to make sure that when some contacts are ++++++ 2015-05-13 22:03 GMT+08:00 Grégory Starck [email protected]:
|
I have more tested the PR in our environment, and it shows some wrong under some situation. I don't know why it will throw an exception when deleting notificationways with an |
I am crazy about the memory linear up. With the patch i give in the pull request, i think i have resolved the memory leak problem. However, we run the modified version with this pull request after four days and the memory actually is still rising. That is very disappointed. |
Is the schedulers memory raising the same or not? (like with comments or On Tue, May 19, 2015 at 11:50 AM, Ning Xie [email protected] wrote:
|
Yes, scheduler seems to be in memory leak too. with the initial memory However, regenerator leaks in WebUI seems has nothing to do with memory Please help me get out of this. I have worked on this for more than a week. ++++++ 2015-05-19 19:07 GMT+08:00 Gabès Jean [email protected]:
|
Ok, so the problem is not in the regenerator or such things I think, nor in If scheduler are impacted, it means that it's upstream the broker, and this Can you try without retention module? Are the memory grow immediate after thanks, On Tue, May 19, 2015 at 1:53 PM, Ning Xie [email protected] wrote:
|
I have not try scheduler without retention module disabled. I will give a I think you can reopen this issue to make more people see it and then may ++++++ 2015-05-19 22:52 GMT+08:00 Gabès Jean [email protected]:
|
it's reopen On Wed, May 20, 2015 at 3:29 AM, Ning Xie [email protected] wrote:
|
I may have some new results about this (about memory leak more exactly) .. |
@gst, It has been a long time ago, i have try to figure it out with more than two weeks. However, the PR now may not work now and i can not remember exactly that whether i worked it out at that time. So, it is difficult for me to recall for this one and i am out of this for more than fourth months. |
This fix the problem issued in #1597 and WebUI 128.