mirror of
https://github.com/kemko/nomad.git
synced 2026-01-01 16:05:42 +03:00
048c5bcba9
* Use core ID when selecting cores
If the available cores are not a continuous set, the core selector might
panic when trying to select cores.
For example, consider a scenario where the available cores for the selector are the following:
[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47]
This list contains 46 cores, because cores with IDs 0 and 24 are not
included in the list
Before this patch, if we requested 46 cores, the selector would panic
trying to access the item with index 46 in `cs.topology.Cores`.
This patch changes the selector to use the core ID instead when looking
for a core inside `cs.topology.Cores`. This prevents an out of bounds
access that was causing the panic.
Note: The patch is straightforward with the change. Perhaps a better
long-term solution would be to restructure the `numalib.Topology.Cores`
field to be a `map[ID]Core`, but that is a much larger change that is
more difficult to land. Also, the amount of cores in our case is
small—at most 192—so a search won't have any noticeable impact.
* Add changelog entry
* Build list of IDs inline
4 lines
171 B
Plaintext
4 lines
171 B
Plaintext
```release-note:bug
|
|
scheduler: Use core ID when selecting cores. This fixes a panic in the scheduler when the `reservable_cores` is not a contiguous list of core IDs.
|
|
```
|