Runaway error messages

[Chris Williams]

We have a situation where a low-traffic OB thread will go into an error condition that fills both log files with the same message at a rate of about 160 messages per second. They all say “[pdl :PDL :ERR /O:  OB_RAW_HCN] read returned error 238 (Connection timed out)“. I’ve been told that “connection timed out” may not necessarily be the definition of error 238. Left unchecked the logs will fill the disk space and the engine shuts down. I have an alert that will cycle the logs, but that’s just a band-aid. This thing is so fast and unending it is as if the engine were in an endless loop. The cpu useage hits around 98% and our users start complaining about system response time. We’re running 5.2.1 on HP-UX 11i. Has anyone else experienced this?

Topic

[Kurt Ciabattari]

Chris,

Every once in awhile I have the same issue with one of our vendors.

One thing you can do is in the network config – configure the process (for the troublesome thread) by checking the box “AUTOMATIC LOG CYCLE” and enter a threshold in kb – -this will make sure that the log files only get so big(threshold) then overwrite themselves and you shouldn’t have to worry about running out of disk space.

As for performance, you may want to try to write an alert that will reset the thread based upon last recieved/send.  I did this and it stopped the error from occuring – -however the external system connection had to be buonced as well.

hope this helps some

good luck

kurt

[Jay Yeiser]

Perhaps you should try another PDL, such as the standard mlp_tcp.  I only use the raw ack for internal engine interfaces.

[Author]

Replies