[Hpc-forum] defected nodes

[Ferenc Bartha]

Dear Users,

There are jobs, which terminate but leave "orphan" processes on the compute 
nodes. SLURM does not know them; launches new jobs without cleaning up. The 
orphan processes damage the new job, sometimes seriously.

These nodes can be cured with using root privileges. As long as NIIF does 
not do it I provide a hint on how to exclude the infected nodes.

Budapest, Budapest2, Debrecen, Debrecen2:
Refer to the about command. It is located in 
/opt/nce/packages/global/barthaf/bin. You can make an alias to it, or simply 
copy it to your own bin, or/and several other solutions to make it easier to 
use...
If you see an "orp" info topic, you'd better take it seriously. I update 
these files typically daily. Remark: For the moment only DB2 is infected, 
but badly.

Pecs, Miskolc:
UV machines are not inspected.

Szeged:
No need for info on this topic, I clean up immediately when orphans are 
recognized.

ÜdvFeri
------------:
Dr. Ferenc BARTHA, tel: 62/54-6821, E-mail: barthaf at sol.cc.u-szeged.hu
SKYPE: ferenc.bartha, WWW: http://www.staff.u-szeged.hu/~barthaf/
SZTE DNT - High Performance Computing Group
   6725 Szeged, Tisza Lajos krt. 113 (Szikra u. 2.)
SZTE, Department of Medical Chemistry, 6720 Szeged, Dóm tér 8.