Hans Lehmann(deleted)
|
system load problem caused by socket
|
Hans Lehmann(deleted)
10/31/2016 9:14 AM
post117033
|
system load problem caused by socket
Hello,
We had a strange problem with the system load which was caused be a socket at a customer. The Socket runs on priority 23
and there is no other process running at this priority.
The socket is used for phindows sessions, network printing and remote service via Phindows.
The time went wrong for some seconds within a minute produced by the system load (10-15%) of the socket.
At this moment we try to analyze the problem there was no print job and only our phindows session established over the
socket which was checked by netstat and the system load was checked by sac. We checked the system load directly on the
node and from another node (sac –n). Furthermore we killed the phindows session to check whether phindows is the reason
or not.
We figured out that the socket causes this problem, because first we restarted several processes without any results and
secondly there is no other process running on priority 23 as mentioned above. After restart of each process we checked
the system time and saw that the time was wrong within a minute and the system load was to high.
After restarting the socket the problem gone and the time is since that date ok and the system works proper without any
system load or problems.
We used Socket Version 4.25H on all machines at the customer but only on some machines this problem occurred.
Do you have any ideas what could cause such a strange problem?
Could several broadcast messages from other systems causes this kind of problems?
Do you have an idea why not all systems are concerned?
Do you have some tools beside netinfo, netsniff, netmap or netstat to analyze this kind of problem?
Kendly reagrds
Hans Lehmann
|
|
|