Umbraco memory leak? Umbraco randomly consumes almost all the RAM on Webserver - requires recycling of app pool
Hi,
I have a site, where it seems to randomly spike up to consume all available RAM on the webserver it runs on.
It stops processing any requests, so the site is down for the users trying to access the site. Any other site being hosted on the same server, is also down, because of this one site spiking.
I have observed it happening at 3.02 in the night, and at 15.01 in the afternoon.
It is running (Umbraco version 7.4.1 assembly: 1.0.5891.23238).
It reminds me of something like a memory leak, but i have no idea as to why this is happening.
It started happening 2-3 weeks ago, and have been happening 2-3 times per week.
The SQL Server is not spiking in requests during the spike of the Umbraco site, so i assume its not a big amount of SQL calls its trying to do.
Have anyone experienced anything similar to this? And perhaps have a solution?
There is multiple Umbraco sites running on this webserver, that doesnt do this stunt. The problematic one is the biggest one of them though. But it is only around 100k nodes.
I've tried this. It works in the sense that it will successfully limit the memory usage. It is however at the cost, that the site does not work at all. It works in the beginning of a new recycled app pool, but once the memory usage climbs, it hits a hard cap of the 32 bit application limit, and then everything just stalls.
There is 7Gb on the server it is running on atm. We have seen it taking up as much as 2,5-2,8Gb and still being somewhat responsive.
Past that point it usually stops and freezes up completely up untill around 4Gb and then it drops to below 100mb and starts climbing in RAM again, while still not processing requests.
I am running into a similar issue. in 64 bits the memory seems to be going up and up until it runs out. We are running a web farm with 4 servers (one of them is used for back office) and eventually they all run out.
We are currently using 7.5.10.
We tried load testing and that seems to be working fine.
It seems to be related to updating content while the server is under load.
We did a process dump and the memory seems to contain the cache in memory. It is almost like the server tries to update the cache when the content has been updated but does not let go of the existing objects.
We tried to change the processes to 32 bits. The CPU got up but also the cache would get corrupted when we update the content and crash the site.
We have other applications which seem to be doing fine. It seems to be related to the number of documents in the application and how often we update the content.
We tried to log a support call but we do not seem to be getting any answer.
At the moment we end up having to recycle the applications 2-3 times a day. That is pretty bad.
I would be curious if we are the only ones having those kind of issues.
Did you ever get this figured out?
We also schedule the app pool recycle, 3 times a day. We do have thousands of documents with daily back-office editing (during business hours).
The site usually crashes during the week, but when checking the logs, it rarely crashes on the weekends. Leading me to believe that this has to be back-office related.
We had to do some process dumps to find out what was happening.
We found a third party library was not behaving and preventing Umbraco from releasing the memory.
Umbraco seems to be restarting a few things whenever you publish documents. All the objects are removed and the cache gets reloaded in memory. We had multiple copies of all the documents in memory, hence the memory leak.
Once we found the issue we were able to fix the issue.
Umbraco support helped us find the actual class which was holding all the objects, after that we were able to fix it.
Umbraco memory leak? Umbraco randomly consumes almost all the RAM on Webserver - requires recycling of app pool
Hi, I have a site, where it seems to randomly spike up to consume all available RAM on the webserver it runs on.
It stops processing any requests, so the site is down for the users trying to access the site. Any other site being hosted on the same server, is also down, because of this one site spiking.
I have observed it happening at 3.02 in the night, and at 15.01 in the afternoon.
It is running
(Umbraco version 7.4.1 assembly: 1.0.5891.23238).It reminds me of something like a memory leak, but i have no idea as to why this is happening.
It started happening 2-3 weeks ago, and have been happening 2-3 times per week. The SQL Server is not spiking in requests during the spike of the Umbraco site, so i assume its not a big amount of SQL calls its trying to do.
Have anyone experienced anything similar to this? And perhaps have a solution?
There is multiple Umbraco sites running on this webserver, that doesnt do this stunt. The problematic one is the biggest one of them though. But it is only around 100k nodes.
We had an issue like this a wee while back and it was because we had an old version of the Url Rewrite plugin loaded into IIS.
We were running a 4.7 version of Umbraco and a pretty old version of Windows so I doubt it's the same issue.
I would check the windows event log and see if it shows any critical issues in there.
Daniel, did you ever find any resolve to your issue? I'm having the exact same problem as you.
No, we haven't resolved it yet. We are planning on trying to get rid of all 3rd party packages though, just to eliminate that possibility.
David,
This was a long time ago and I am experiencing a similar issue now. Did you ever resolve this?
check this: http://lukealderton.com/blog/posts/2015/july/reducing-umbraco-memory-usage-by-enabling-32-bit-applications.aspx
I've tried this. It works in the sense that it will successfully limit the memory usage. It is however at the cost, that the site does not work at all. It works in the beginning of a new recycled app pool, but once the memory usage climbs, it hits a hard cap of the 32 bit application limit, and then everything just stalls.
How much memory RAM you have?
There is 7Gb on the server it is running on atm. We have seen it taking up as much as 2,5-2,8Gb and still being somewhat responsive. Past that point it usually stops and freezes up completely up untill around 4Gb and then it drops to below 100mb and starts climbing in RAM again, while still not processing requests.
I am running into a similar issue. in 64 bits the memory seems to be going up and up until it runs out. We are running a web farm with 4 servers (one of them is used for back office) and eventually they all run out. We are currently using 7.5.10. We tried load testing and that seems to be working fine. It seems to be related to updating content while the server is under load.
We did a process dump and the memory seems to contain the cache in memory. It is almost like the server tries to update the cache when the content has been updated but does not let go of the existing objects.
We tried to change the processes to 32 bits. The CPU got up but also the cache would get corrupted when we update the content and crash the site.
We have other applications which seem to be doing fine. It seems to be related to the number of documents in the application and how often we update the content.
We tried to log a support call but we do not seem to be getting any answer. At the moment we end up having to recycle the applications 2-3 times a day. That is pretty bad.
I would be curious if we are the only ones having those kind of issues.
Yann,
Did you ever get this figured out? We also schedule the app pool recycle, 3 times a day. We do have thousands of documents with daily back-office editing (during business hours).
The site usually crashes during the week, but when checking the logs, it rarely crashes on the weekends. Leading me to believe that this has to be back-office related.
We had to do some process dumps to find out what was happening. We found a third party library was not behaving and preventing Umbraco from releasing the memory. Umbraco seems to be restarting a few things whenever you publish documents. All the objects are removed and the cache gets reloaded in memory. We had multiple copies of all the documents in memory, hence the memory leak.
Once we found the issue we were able to fix the issue. Umbraco support helped us find the actual class which was holding all the objects, after that we were able to fix it.
Do you know what library this was? I have had suspicious of one that we use, Archetype, but I haven't been able to pin-point any solid evidence.
For us it was related to using RabbitMQ message queue. Sorry, it is probably not that helpful.
is working on a reply...