[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20051208: Weird LDM occurrence

>From:  Mark Bradford <address@hidden>
>Organization:  NCAR/EOL/CDS/CTM
>Keywords:  200512081637.jB8Gbe7s009853 LDM assertion error

Hi Mark,

>Tom, last night my LDM exited, with the following message:
>
>Dec 08 05:24:22 torrent thelma[273]: assertion "n > 0" failed: file
>"pq.c", line
> 2187
>Dec 08 05:24:28 torrent rpc.ldmd[267]: child 273 terminated by signal 6
>Dec 08 05:24:28 torrent rpc.ldmd[267]: Killing (SIGINT) process group

You are the fourth person to report this same error at the same time!
Steve E. will dig into the cause when he returns from the AGU meeting
in San Francisco.

>I did a little Googling, and found messages that suggested this could be
>caused by a corrupt queue.  So, even though 'ldmadmin queuecheck'
>returned cleanly, I deleted and rebuilt the queue, and got the same
>error on startup.

This happened to me on a dual Xeon EM64T machine running LDM-6.4.2 
yesterday as well.  The failure went away after I remade the queue
again.  I am getting the feeling that the failures being reported
had something to do with the time!?

>Next, I noticed that the error was registering on a particular ldmd; I
>have two that point to thelma, one of which uses a more restrictive
>regexp for requesting.  The one that was erroring had the line:
>request "PPS|DDS|WMO|FSL2"   ".*"    thelma.ucar.edu

Please change all feed requests aimed at thelma.ucar.edu to
idd.unidata.ucar.edu.  'thelma.ucar.edu' is currently an alias for
idd.unidata.ucar.edu we setup when we decommissioned the Sun
SPARC-based thelma several months ago.

>"Aha," I thought, "I'll split out that regexp and see which specific
>feedtype has the problem."  So I turned it into four separate "request"
>lines (even though I know you prefer the combined one)... and then LDM
>started with no error.

I think that the things started working not because of a change
in feed requests, but, rather, because it was later.

>So that's the current state.  Anything I should do from here?

Please change your feed requests back to the way they were before
(changing the machine from 'thelma.ucar.edu' to idd.unidata.ucar.edu)
and let us know if you see any problems.

>Thanks!

No worries.  This is a very strange problem indeed.

Cheers,

Tom

>From address@hidden  Thu Dec  8 12:02:34 2005

Tom Yoksas wrote:

re: your feed requests back to the way they were before

>Done.  So far, so good...

>-- 
>Mark Bradford, Group Head, Collaborative Technologies and Metadata
>NCAR Earth Observing Laboratory/CDS <> address@hidden <> (303) 497-8169