Log Slurm topology node lists as a single nodes_csv column (#177) - #177
Merged
Conversation
|
@mitthu has exported this pull request. If you are a Meta employee, you can view the originating Diff in D118152543. |
CI CommandsThe following CI workflows run automatically on every push and pull request:
The following commands can be used by maintainers to trigger additional tests that require access to secrets:
|
luccabb
approved these changes
Aug 31, 2026
luccabb
left a comment
luccabb
left a comment
Contributor
There was a problem hiding this comment.
this is good, we should maybe patch creation at some point: https://github.com/facebookresearch/gcm/blob/main/gcm/monitoring/dataclass_utils.py#L103
Comment on lines
+30
to
+32
| # Comma-separated form of `Nodes`. Sinks flatten list fields into one | ||
| # indexed column per element (`Nodes.0`, `Nodes.1`, ...), which makes the | ||
| # node list unusable as a single value in Scuba. This keeps it queryable. |
Contributor
There was a problem hiding this comment.
Suggested change
| # Comma-separated form of `Nodes`. Sinks flatten list fields into one | |
| # indexed column per element (`Nodes.0`, `Nodes.1`, ...), which makes the | |
| # node list unusable as a single value in Scuba. This keeps it queryable. |
Contributor
Author
There was a problem hiding this comment.
Updated to just flatten the source list instead of adding the nodes_csv column.
meta-codesync
Bot
changed the title
nodes_csv columnnodes_csv column (#177)
mitthu
added a commit
to mitthu/gcm
that referenced
this pull request
Aug 31, 2026
…kresearch#177) Summary: `ScontrolTopology.Nodes` is a `list[str]`, and the sinks flatten list fields into one indexed attribute per element (`flatten_dict_factory` in `monitoring/dataclass_utils.py`, used by `exporters/otel.py`). In Scuba that turns a single node list into 1182 separate `Normals` columns, `Nodes.0` through `Nodes.1181` — with no array left to `IMPLODE`, and a new column minted every time a switch grows. Getting the node list back as one value today requires a derived column that enumerates every index: ``` IMPLODE(',', REGEXP_REMOVE_ARRAY(ARRAY(`Nodes.0`, ..., `Nodes.1181`), '^$')) ``` That is ~15KB of SQL, reads 1182 columns per row, and silently truncates the moment a node list outgrows the highest column that exists. This adds a `nodes_csv` field alongside `Nodes` holding the comma-joined list, which lands as a single `Normals` column — groupable and filterable with `substr`/regex, no derived column needed. `Nodes` is kept as-is so the per-element columns remain available and ordering is preserved (a tagset would dedupe and reorder, which would break the topology-aware job debugging use case in T275974051). Empty node lists (e.g. `SwitchName=data-transfer Nodes=`) leave `nodes_csv` unset rather than emitting an empty string, matching how the other optional fields behave. Differential Revision: D118152543
mitthu
force-pushed
the
export-D118152543
branch
from
August 31, 2026 19:07
852830c to
2332832
Compare
…kresearch#177) Summary: `ScontrolTopology.Nodes` is a `list[str]`, and the sinks flatten list fields into one indexed attribute per element (`flatten_dict_factory` in `monitoring/dataclass_utils.py`, used by `exporters/otel.py`). In Scuba that turns a single node list into 1182 separate `Normals` columns, `Nodes.0` through `Nodes.1181` — with no array left to `IMPLODE`, and a new column minted every time a switch grows. Getting the node list back as one value today requires a derived column that enumerates every index: ``` IMPLODE(',', REGEXP_REMOVE_ARRAY(ARRAY(`Nodes.0`, ..., `Nodes.1181`), '^$')) ``` That is ~15KB of SQL, reads 1182 columns per row, and silently truncates the moment a node list outgrows the highest column that exists. This adds a `nodes_csv` field alongside `Nodes` holding the comma-joined list, which lands as a single `Normals` column — groupable and filterable with `substr`/regex, no derived column needed. `Nodes` is kept as-is so the per-element columns remain available and ordering is preserved (a tagset would dedupe and reorder, which would break the topology-aware job debugging use case in T275974051). Empty node lists (e.g. `SwitchName=data-transfer Nodes=`) leave `nodes_csv` unset rather than emitting an empty string, matching how the other optional fields behave. Differential Revision: D118152543
mitthu
force-pushed
the
export-D118152543
branch
from
August 31, 2026 19:13
2332832 to
462a773
Compare
Yash0270
approved these changes
Aug 31, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary:
ScontrolTopology.Nodeswas alist[str], and the sinks flatten list fieldsinto one indexed attribute per element (
flatten_dict_factoryinmonitoring/dataclass_utils.py, used byexporters/otel.py). In Scuba thatturned a single node list into 1182 separate
Normalscolumns,Nodes.0through
Nodes.1181, with a new column minted every time a switch grows.Differential Revision: D118152543