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

[TIGGE #EEC-237791]: Memory problem with LDM Server at CMA

MA Qiang,

> The syslog.conf file was overwriten wrongly by CSM cfmupdatenoed command, and 
> I just corrected it before you logged in but I did not flush the syslogd. 
> Sorry to confused you.  Today, when I stopped LDM and wanted to umount the 
> GPFS filesystem, the system hung. So I restart the LDM server again. Now the 
> ldmd.log is logged.

Good.  The LDM log messages might help.

> 　　　The LDM product-queue is on a disk that is local to tgn01. This 
> disk(/dev/mapper/vg_root-lv_ldm_pq, /space/ldm_pq) can not be accessed from 
> another computer via GPFS.
> The swap-file for tgn01 is also on a local disk and can not be accessed from 
> another computer via GPFS.

Good.

I recommend that you increase the size of your LDM product-queue,
if possible.  The output of "pqmon"

    $ pqmon -q /space/ldm_pq/ldm.pq
    Jan 16 22:17:17 pqmon NOTE: Starting Up (32004)
    Jan 16 22:17:17 pqmon NOTE: nprods nfree  nempty      nbytes  maxprods  
maxfree  minempty    maxext  age
    Jan 16 22:17:17 pqmon NOTE:   7029  1509  616462  2437516304     57717     
4704    566504    428800 641
    Jan 16 22:17:17 pqmon NOTE: Exiting

shows that the oldest data-product in the queue was inserted into
the queue only 641 seconds ago.  Such a small queue can result in
warning messages like

    Jan 16 22:03:17 tgn01 pqact[15984] WARN: Processed oldest product in queue: 
628.452 s 

and risks "pqact" not processing all data-products.  Normally,
we advise that the product-queue hold about one hour's worth of
data.  Unfortunately, increasing the size of the queue will
increase the size of the LDM processes and might cause your
problem to become worse.  Can you increase the amount of memory?

Regards,
Steve Emmerson

Ticket Details
===================
Ticket ID: EEC-237791
Department: Support IDD TIGGE
Priority: Normal
Status: On Hold