RichardDCG Posted September 7 Report Share Posted September 7 I have increased the RAM from 2GB to 4GB on our Vodia PBX but still getting faults registered. The last one below is 1 single active call on the pbx. Are these graphs accurate? We are not experiencing any call quality issues, just going off the notes below to increase RAM and not seeing any change... "It is okay if this happens occasionally; however if there are many page faults the system should have more memory. " Quote Link to comment Share on other sites More sharing options...
Vodia PBX Posted September 7 Report Share Posted September 7 The graph comes from what the operating system tells the process in terms of page faults. I would say they are accurate. The graph is a call for action to check out what is going on. Paging is a complex topic. I spent some time but did not find a simple way of telling the OS (Linux) to disable paging for a process, but did not really find anything (maybe someone else has an idea, please post it). We had a place in the code where we were using mlockall, but it's not accessible. We'll add an option --paging false. But nothing comes for free. When you disable paging, the PBX might run out of memory when otherwise paging would still keep the process running. Quote Link to comment Share on other sites More sharing options...
RichardDCG Posted September 8 Author Report Share Posted September 8 If a single call will raise a page fault, how much memory should I have? All other graphs seem fine (memory, CPU etc.) Voice quality also seems to be good. Quote Link to comment Share on other sites More sharing options...
Vodia PBX Posted September 9 Report Share Posted September 9 We did not have the page fault graph before and in most cases there is no need to worry about it. But we had cases where the PBX got constantly swapped out and we were not really aware about it. That's why we added it. If you want to provide a great service, those early warning signs are important. Looking at other graphs, usually there is maybe one or two page faults in a day. But in your case, you have page faults distributed all over the day. This does raise the question what is going on. I don't even know what OS this is. Maybe Windows puts more memory pressure on the PBX? It's definitely an early warning because the call quality seems to be okay. But still it's worth investigating why the OS feels like it has to swap pages in and out of the PBX process. Maybe time to search more for telling the OS that the PBX should not have any memory pressure? Quote Link to comment Share on other sites More sharing options...
RichardDCG Posted September 9 Author Report Share Posted September 9 25 minutes ago, Vodia PBX said: Maybe Windows puts more memory pressure on the PBX This is a Linux server. It does seem to be usnig some swap ... Quote Link to comment Share on other sites More sharing options...
Vodia PBX Posted September 19 Report Share Posted September 19 The PBX is polling getrusage(2) and looks at ru_majflt. Every time there is a change, it adds one tick to the pagefault counter. The majflt is "the number of page faults serviced that required I/O activity." I believe we are looking at the right number. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.