Editing Routine Maintenance (section)

== check load graphs == 

Click on the Load link in mgmt 

This screen shows you load levels on our servers and network traffic for critical machines (firewalls, backup servers).

If you see load high or increasing 

FreeBSD: 
run [[VPS_Management#jtop|jtop]] (or [[VPS_Management#jt|jt]] > 7.x) and see if there are any runaway processes.  Here are some examples of entries in top that are
definitely runaway processes:

<pre>79481 root      64   0  2256K  1056K CPU1   1  58:16 87.40% 87.40% nano
50650   1000    64   0  1852K  1112K RUN    0 207.9H 84.08% 84.08% screen
14829 www        2   0 39100K 31736K accept 0  104:24  46.54%  6.54% httpd
42065 root      61   0  1300K   844K RUN    1  47.8H 91.36% 91.36% ee
1328 www       56   0 18440K 10796K CPU1   0  64.4H 97.71% 97.71% httpd
26251 user      57   0  6124K  1160K CPU1   1  82.9H 98.44% 98.44% screen
89874 root      60   0  1352K   892K RUN    1  33.8H 65.82% 65.82% dialog
38656   1000    64   0  3088K  2136K CPU0   0 806:13 97.95% 97.95% StutBot
27630 root      64   0  1396K   972K RUN    1  76.8H 86.47% 86.47% ee</pre>

Linux:
run [[VPS_Management#vwe|vwe]] to see which VPS’s have high loads. From there run <tt>[[VPS_Management#vp|vp]] <veid></tt> and/or <tt>[[VPS_Management#vt|vt]] <veid></tt> to see what's going on in that system.
[[VPS_Management#vzstat|vzstat]] will also give you a nice picture of whats going on, systems with high numbers in the mlat column are likely culprits.

examples of out of control procs:

<pre>12183 nobody    16   0  4916 1348  1340 R    45.5  0.0  4249m httpd
29266 #502      16   0  1852  796   792 R    22.5  0.0  1104m vim
23860 #41       16   0  5472 5472  2076 R    98.9  0.2  31:41 python
19227 bin       19   0  1688  716   652 R    99.9  0.0 321:08 wtrs_ui
 7762 apache    16   0   268  236   224 R    85.7  0.0  1010m ptrace
 4624 #501      20   0  4304 2400  2044 R    53.6  0.1 284:32 YoSucker
20451 #506      20   0  1876  820   816 R    17.2  0.0 169:35 vim
 8834 #514      20   0   900  724   672 R    77.6  0.0 382:30 neostats
31815 apache    14   0  3176 3176  1696 R    74.4  0.1   6:15 counter</pre>

Just kill -9 them and be done with it.

Also, anytime you see `kmod` or `ptrace` - kill those immediaely no matter how much they are using - they are attempts to exploit the linux ptrace bug.  They won't work, but they suck a lot of CPU...

Also, any other processes that are at 90-100% cpu usage and have been running for any long period of time should be killed except for mysqld processes on FreeBSD.  See above.

However, there is an exception:

if it is a mysqld, we don't want to kill their database.  What you want to do is <tt>[[VPS_Management#jpid|jpid]] <pid></tt> to see who owns it, and then email them the paste containing the instructions for the nanny. Or you can simply do a <tt>kill -1 PID</tt> on the process to restart it.

=== Load averages jump at night ===

The load averages on the FreeBSD systems may jump up at night between 1 and 4 am - this is because the backups are running - if this is what is causing a jump in load, you will see processes like `rsync` in top eating a lot of CPU time.