Hello all,
I've recently been experimenting with dmq, and I've noticed something that seemed a little odd. shmem use would often be abnormally high on otherwise idle kamailio nodes (literally, only dispatcher module OPTIONS were exchanged at the time besides KDMQ messages).
After disabling the "enable_dmq" parameter for the dialog module, this sort of absurd behaviour ceased. To get a quantified idea of this change's effect, check out the following screenshot from our monitoring system: https://pasteboard.co/Jb430lD.png
As you can see, the spikes stopped manifesting at around 05:00, when dmq was disabled for dialog (albeit still employed for htable). shmem use then grew as traffic was introduced to these nodes, and htables began filling with data.
I was wondering if this to be expected, which I find highly odd given that at the time other productive systems in the DMQ cluster were also idle (given the time of the day), not to mention that their shmem usage was not nearly as high.
Could this be attributed to some configuration error, with updates looping endlessly (or until some counter reaches zero or something) among nodes? Or could memory be used for other housekeeping purposes unrelated to dmq dialog sharing?
Lastly, what is the meaning (and differences) of these shmem stats?
- real_used_size - used_size - max_used_size
Max used size I get (I'm guessing maximum seen since server was started?) but what's the difference between real_used and used? Thanks!
Best regards, George
Hello,
I get image not found on that link, have you removed it?
Cheers, Daniel
On 01.06.20 13:57, George Diamantopoulos wrote:
Hello all,
I've recently been experimenting with dmq, and I've noticed something that seemed a little odd. shmem use would often be abnormally high on otherwise idle kamailio nodes (literally, only dispatcher module OPTIONS were exchanged at the time besides KDMQ messages).
After disabling the "enable_dmq" parameter for the dialog module, this sort of absurd behaviour ceased. To get a quantified idea of this change's effect, check out the following screenshot from our monitoring system: https://pasteboard.co/Jb430lD.png
As you can see, the spikes stopped manifesting at around 05:00, when dmq was disabled for dialog (albeit still employed for htable). shmem use then grew as traffic was introduced to these nodes, and htables began filling with data.
I was wondering if this to be expected, which I find highly odd given that at the time other productive systems in the DMQ cluster were also idle (given the time of the day), not to mention that their shmem usage was not nearly as high.
Could this be attributed to some configuration error, with updates looping endlessly (or until some counter reaches zero or something) among nodes? Or could memory be used for other housekeeping purposes unrelated to dmq dialog sharing?
Lastly, what is the meaning (and differences) of these shmem stats?
- real_used_size
- used_size
- max_used_size
Max used size I get (I'm guessing maximum seen since server was started?) but what's the difference between real_used and used? Thanks!
Best regards, George
Kamailio (SER) - Users Mailing List sr-users@lists.kamailio.org https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users
Nope, it opens fine here. Does this work: https://snipboard.io/5E4V3m.jpg ?
On Tue, 2 Jun 2020 at 08:34, Daniel-Constantin Mierla miconda@gmail.com wrote:
Hello,
I get image not found on that link, have you removed it?
Cheers, Daniel On 01.06.20 13:57, George Diamantopoulos wrote:
Hello all,
I've recently been experimenting with dmq, and I've noticed something that seemed a little odd. shmem use would often be abnormally high on otherwise idle kamailio nodes (literally, only dispatcher module OPTIONS were exchanged at the time besides KDMQ messages).
After disabling the "enable_dmq" parameter for the dialog module, this sort of absurd behaviour ceased. To get a quantified idea of this change's effect, check out the following screenshot from our monitoring system: https://pasteboard.co/Jb430lD.png
As you can see, the spikes stopped manifesting at around 05:00, when dmq was disabled for dialog (albeit still employed for htable). shmem use then grew as traffic was introduced to these nodes, and htables began filling with data.
I was wondering if this to be expected, which I find highly odd given that at the time other productive systems in the DMQ cluster were also idle (given the time of the day), not to mention that their shmem usage was not nearly as high.
Could this be attributed to some configuration error, with updates looping endlessly (or until some counter reaches zero or something) among nodes? Or could memory be used for other housekeeping purposes unrelated to dmq dialog sharing?
Lastly, what is the meaning (and differences) of these shmem stats?
- real_used_size
- used_size
- max_used_size
Max used size I get (I'm guessing maximum seen since server was started?) but what's the difference between real_used and used? Thanks!
Best regards, George
Kamailio (SER) - Users Mailing Listsr-users@lists.kamailio.orghttps://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users
-- Daniel-Constantin Mierla -- www.asipto.comwww.twitter.com/miconda -- www.linkedin.com/in/miconda Funding: https://www.paypal.me/dcmierla
-
Daniel-Constantin Mierla -
George Diamantopoulos