rewb0rn wrote:Yes it is the extension queue. However the problem only appears on 2.13.0 and 2.13.1 (2.13.2 only installed yesterday). The last 3 months we were switching back to 2.11.0 and the problem is not happening there. But we are publishing to HTML5 now so we need binary wss protocol.
Okay. I don't think there's anything specific in 2.11 compared to 2.13 that changes how queue works. In fact that aspect of the server has been the same for a long time (since 2.9.x I believe).
It seems like there's a lot of pressure on the Extension queue, due to those slow I/O operations. Unfortunately the only "cure" for that is parallelism, which in turn means more threads, which in turn require more memory (each thread has his own stack memory).
Using thread pools alleviates the memory issue greatly.
If you're running 1000s of threads and have high traffic (i.e. 10K as you mentioned) you will probably need 4-8Gb of RAM for SFS2X alone, though I would not worry too much about memory if no OOM have been seen in the logs.
I was thinking that our backup settings were too sensitive, so I switched backup triggering time from 3 seconds to 20 seconds, but 2 days ago it happened again.
Yeah, 3 seconds seems excessive, I'd probably use the 15-20 second is already pretty low.
Code: Select all
By the way: Is there a way to read the values that are tracked in the dashboard through code? I would like to have a better record of statistics and write them to a file instead of having the dashboard open in another tab.
Depends on which values you need to read.
Send us an email to our support@... email box with the details and we'll try to help you further on this topic.
(add a link to this thread for reference)
Cheers