Bug #9704
closedAs for Rudder 3.2.9, promises calculation is still too slow
Description
At this moment, rudder server processes hosts one-by-one, independently from grouping . In my case, calculations for 1 host takes 5…6 seconds, so 80 hosts in group are processed about 6 minutes. (But less then 2 CPU are used at each point of time.)
This is the ceiling latency. But there can be about 10'th times more hosts in management in case of implementing parallel calculations.
What I expect: as we know the list of hosts with the same set of directives, there can be scheduled N parallel sub-tasks, where N must be set in config, depending on number of CPU's on server host.
The other way is to design "make one — clone to many": make calculations for the group template, then use it to write hosts-specific bundles for each host in group.
- The other side of performance is HDD speed, but it is to be improved at the server-side. Using ssd can speed-up in about 1 minute from 6.
Updated by François ARMAND over 8 years ago
- Related to User story #5641: Make the agent policies update a state machine with integrity check added
Updated by François ARMAND over 8 years ago
- Related to User story #8353: Implement notifications for different server-side actions and events (hooks) added
Updated by Benoît PECCATTE over 8 years ago
- Severity set to Minor - inconvenience | misleading | easy workaround
- User visibility set to Getting started - demo | first install | level 1 Techniques
- Priority set to 0