-
Created by Unknown User (robyn), last modified on May 13, 2014
Date
May 13, 2014
Attendees
Discussion Items
Name | Notes |
---|
K-T | - All tasks through August need to be put into JIRA by tomorrow afternoon
|
Steve | Steve working on distributor stuff - Need permanent replicator/distributor connection instead of dynamic
- See up to 20 seconds to start replicator jobs
- Can probably adjust negotiator parameters to reduce
|
Greg | - Still working on making "stuck job" scenario work more optimally
- Cascade effect; job bumps another cache, doesn't use spare slot
|
Condor Team | - Dynamic ClassAds don't seem to be able to set slot-based parameters
+ Even for static slots (as opposed to partitionable slots) + startd_cron variables seem to be set for all slots + Todd will investigate- Could do attribute across slots on node
+ With partitionable slots, can do custom resource binding + Probably need to have locking, reference counting of the cache- CMS has time to work on previously-circulated cache design
- schedd can be told to keep claim for faster job startup
|
Action Items