Hi,
I’m using Pulse to start and shut down slaves on demand, however Pulse seems to be starting slaves for jobs even though they’re not in the jobs’ pool. As a test, I installed Maya on one node and added it to a new pool. When I submitted the Maya job, all the machines were started (and kept starting up when I shut them down) while the job was active. The priority structure is set to the default pool_priority_date but that shouldn’t really matter anyway.
There is also a default pool called ‘none’ that I can’t remove now there are other pools to choose from.
Cheers,
Jon
Hi Jon,
I can’t seem to reproduce this with our current version. Are you using 3.0 or 3.0 SP1?
I created a new pool called test_pool and assigned it to one slave. I then submitted the job and selected ‘test_pool’ as the pool and ‘none’ as the group. The only machine that was woken up was the single slave that I had assigned ‘test_pool’ to.
When Pulse is performing the power management check, it outputs information to the Pulse log and to the appropriate section under the Power Management tab in Pulse. Can you post what it is printing out when it’s performing the Wake On Lan check?
Thanks,
Hi,
I’ll test that out in a little while once we finish a software upgrade on the farm. Just to check, we’re running version 3.0.33353 which is named as Deadline 3.0, however I definitely installed SP1 when I upgraded the render farm so perhaps it doesn’t list it as such.
Thanks,
Jon
Hi Ryan,
I performed a test earlier and received the following information under the Power Management tab:
Checking job “mayatest_01” (004_050_999_7af683f6)
- job is not part of a pool, any slaves should be woken up
- slaves that are candidates to be woken up: 102-Rdr,105-Rdr,107-Rdr,106-Rdr,103-Rdr,117-Rdr,108-Rdr,116-Rdr,113-Rdr,104-Rdr,101-Rdr,118-Rdr,100-Rdr,111-Rdr,109-Rdr,110-Rdr,112-Rdr,119-Rdr
- checking slave 102-Rdr
- waking up offline slave 102-Rdr because it is required to render job “mayatest_01”
- checking slave 105-Rdr
- waking up offline slave 105-Rdr because it is required to render job “mayatest_01”
- checking slave 107-Rdr
- waking up offline slave 107-Rdr because it is required to render job “mayatest_01”
- checking slave 106-Rdr
- waking up offline slave 106-Rdr because it is required to render job “mayatest_01”
- checking slave 103-Rdr
- waking up offline slave 103-Rdr because it is required to render job “mayatest_01”
- checking slave 117-Rdr
- waking up offline slave 117-Rdr because it is required to render job “mayatest_01”
- checking slave 108-Rdr
- waking up offline slave 108-Rdr because it is required to render job “mayatest_01”
- checking slave 116-Rdr
- waking up offline slave 116-Rdr because it is required to render job “mayatest_01”
- checking slave 113-Rdr
- waking up offline slave 113-Rdr because it is required to render job “mayatest_01”
- checking slave 104-Rdr
- waking up offline slave 104-Rdr because it is required to render job “mayatest_01”
It seems to think that the Maya task isn’t part of a pool, even though it has been assigned one and it is selected in the job properties (it’s listed as maya_renderfarm)
Thanks,
Jon
Hi Jon,
Thanks for the log. It looks like this is a bug with the current release of Deadline that has been fixed in our internal working version. We did fix a few things in power management since the last release, and it appears this one also got fixed along the way. The fix will be included in the next release (in a few months), but in the meantime there is unfortunately no workaround, and we apologize for this inconvenience.
Cheers,