Bug 52 - Scheduler locking support
: Scheduler locking support
Status: NEW
Product: TORQUE
pbs_server
: 2.4.x
: PC Linux
: P5 enhancement
Assigned To: Ken Nielson
:
:
:
  Show dependency treegraph
 
Reported: 2010-04-19 03:47 MDT by Simon Toth
Modified: 2010-09-14 22:43 MDT (History)
4 users (show)

See Also:


Attachments
Patch against current 2.4-fixes head (18.16 KB, patch)
2010-04-19 03:47 MDT, Simon Toth
Details | Diff


Note

You need to log in before you can comment on or make changes to this bug.


Description Simon Toth 2010-04-19 03:47:43 MDT
Created an attachment (id=26) [details]
Patch against current 2.4-fixes head

Scheduler locking is a locking technique to prevent several schedulers (or even
one scheduler, when interacting with user tools) connected to one server to
negatively impact each others performance.

New PBS Batch command is implemented with 3 subcommands:
-> acquire lock (fails if another lock is present)
-> refresh lock (refresh the lock timeout)
-> release lock (release the lock)

When a lock is present on the server, PBS batch commands are processed
depending on their category. Commands that are read-only, or work with jobs
that are not in queued state are passed through, commands that affect queued
jobs are either deferred (run immediately when the lock is released), or
blocked.

When no lock is present, the system behaves as if no locking mechanism exists.
Comment 1 Glen 2010-09-14 21:05:20 MDT
it this new feature going to get merged into trunk?  any thoughts from other
developers?
Comment 2 Chris Samuel 2010-09-14 22:43:46 MDT
The idea sounds reasonable enough.