Loomio

Server load

PP Pirate Praveen Public Seen by 20

I think we are hitting the load limits on our server.

diaspora@poddery:~$ free -m
total used free shared buffers cached
Mem: 3965 3816 149 0 4 575
-/+ buffers/cache: 3235 730
Swap: 3355 1256 2099

Stream is loading slow and sometimes nginx gives a page not found message. How do we handle this?

PP

Pirate Praveen Tue 18 Mar 2014 10:14AM

@manukrishnantv , 1256 M swap is unusual I think.

MJ

Minto Joseph Wed 19 Mar 2014 3:00AM

I ran following while loading several streams..

diaspora@poddery:~$ top -b -n 10 >/tmp/top2.out
diaspora@poddery:~$ grep Cpu /tmp/top.out
%Cpu(s): 1.3 us, 0.3 sy, 0.0 ni, 97.2 id, 1.0 wa, 0.0 hi, 0.0 si, 0.1 st
%Cpu(s): 2.7 us, 0.6 sy, 0.0 ni, 96.2 id, 0.3 wa, 0.0 hi, 0.1 si, 0.1 st
%Cpu(s): 2.1 us, 0.5 sy, 0.0 ni, 95.1 id, 2.0 wa, 0.0 hi, 0.0 si, 0.4 st
%Cpu(s): 0.9 us, 0.1 sy, 0.0 ni, 96.9 id, 1.7 wa, 0.0 hi, 0.0 si, 0.3 st
%Cpu(s): 0.1 us, 0.1 sy, 0.0 ni, 99.7 id, 0.0 wa, 0.0 hi, 0.0 si, 0.1 st
%Cpu(s): 3.1 us, 0.4 sy, 0.0 ni, 92.8 id, 3.4 wa, 0.0 hi, 0.1 si, 0.2 st
%Cpu(s): 7.8 us, 1.5 sy, 0.0 ni, 85.6 id, 3.7 wa, 0.0 hi, 0.1 si, 1.3 st
%Cpu(s): 4.2 us, 0.8 sy, 0.0 ni, 92.9 id, 2.0 wa, 0.0 hi, 0.1 si, 0.1 st
%Cpu(s): 8.6 us, 1.2 sy, 0.0 ni, 81.9 id, 4.6 wa, 0.0 hi, 0.1 si, 3.6 st
%Cpu(s): 3.8 us, 1.0 sy, 0.0 ni, 91.8 id, 2.5 wa, 0.0 hi, 0.1 si, 0.8 st
diaspora@poddery:~$ date
Wed Mar 19 08:09:13 IST 2014

If the memory usage was casusing a serious effect, it would have reflected in sys% usage.

stream loading was also fast.

--

Most of the memory looks accounted by processes. Yes, the memory and swap usage seem very high and could result in funny results when lots of users are logged in.

             total       used       free     shared    buffers     cached

Mem: 4061112 3924404 136708 0 4412 606724
-/+ buffers/cache: 3313268 747844
Swap: 3436540 1455928 1980612

Just that it did not translated into an issue when I checked.

--

May be I tested when there were not much users. Is there a particular time when the issue hits?

We might need to install sysstat to check for historical load details.

PP

Poll Created Fri 28 Mar 2014 11:49AM

Stop new sig ups Closed Thu 3 Apr 2014 11:26AM

Outcome
by Pirate Praveen Tue 25 Apr 2017 5:22AM

We have solved the slow stream load issue, no need to close sign ups now.

We cannot provide a good experience with current load. So immediately stop new sign ups.

Then think about moving some of the other services running on this server to their own hosting.

Results

Results Option % of points Voters
Agree 50.0% 2 NAJ JS
Abstain 0.0% 0  
Disagree 50.0% 2 PP MKT
Block 0.0% 0  
Undecided 0% 37 S MK J NV BC P AKS RD S PS AA SAJ AR S JB P J MJ SK FMP

4 of 41 people have participated (9%)

PP

Pirate Praveen
Agree
Fri 28 Mar 2014 11:50AM

yes, we cannot continue like this. This will energize campaign for more pods.

JS

Juan Santiago
Agree
Fri 28 Mar 2014 11:56AM

It is better for the diaspra ecosystem, servers working properly and have not more work than they can bear, then if I find the right opening pisble new records again.

MKT

Manu Krishnan T V
Disagree
Fri 28 Mar 2014 3:18PM

This hasn't caused any serious issues yet.

Also, we are planning to get a new dedicated server by the start of April. Could donate a VM to SMC or FCI to move few of the sites. Hosting sites of SMC in US won't be an issue I believe.

MKT

Manu Krishnan T V
Disagree
Fri 28 Mar 2014 3:54PM

We are planning to get a new dedicated server by the start of April. Could donate a VM to SMC or FCI to move few of the sites. Hosting sites of SMC in US won't be an issue I believe.

PP

Pirate Praveen
Disagree
Thu 3 Apr 2014 11:26AM

the server load issue is resolved and now the experience is smoother again. We can continue with open sign ups.

MKT

Manu Krishnan T V Fri 28 Mar 2014 3:14PM

The memory and swap usage seems high, but I don't find that affecting the overall performance.

It seems like memory is not freed by some processes (expecially php-fpm and ruby) and the system has been running for more than 100 days.

If it starts creating problems, I would suggest restarting all processes and see how the memory usage is.

MKT

Manu Krishnan T V Fri 28 Mar 2014 3:57PM

It seems like the stream loads slow now. The SMC sites was moved to the Poddery server when they had some issues as a temporary solution.

Let me contact Anivar and update the status there. Also, as I mentioned in the proposal, I hope we could donate a VM by mid of April, after all the testing.

Load More