Re: [Condor-users] job's id: cluster.process

["Matt Hope" <matthew.hope@xxxxxxxxx>]

Aswaka jawaka wrote:
> Hi, I was wonderring what does the cluster part of the id mean,

In addition to Todd's statement the following two provisos hold

1) cluster ids are unique to a schedd. Thus they are not an id you can
use as a primary key within a multiple submit host environment, They
are *locally* monotonic and unique...

..Except when
2) you wipe the schedd queue log file. Then subsequent jobs get id's
starting from 0 again.

The first restriction is easy enough to work around since you really
should be using a composite PK (in later versions of condor they
generate a very low collision chance key for the job based on the
schedd name - if you really care put your own GUIDs in a user defined
attribute. The second is not, make sure that anything which uses the
monotonic nature of the ClusterIds to order things (say historical
output) is tolerant to resets (say a new install or a drive crash and
rebuild)

Matt