Details
-
Bug
-
Resolution: Unresolved
-
Minor
-
None
-
None
-
3
-
11540
Description
25.5 - There are several metrics (test operations) that can be collected by mds-survey. In the default execution five (5) are reported: create, lookup, setattr, getattr and destroy - in that order. This "default" info should be added to the first section, along with:
a) list of minimum operations to run (create and destroy),
b) an example of running with the minimum operations
c) an example of running with another supported, but non-default execution (like create, getattr, destroy)
25.5.1 - In the "Note" about not cleaning up files when execution is manually stopped using CNTL-C, please add a section about mounting the target as type ldiskfs, and recursively deleting the "tests*" directories under /mdt/ROOT/ if subsequent executions fail to run.
25.5.2 - The output is truncated. Am not sure what to suggest, as the default produces 5 different metrics (long output for each line). Perhaps run with the minimum operations and let it wrap?
25.5.2 - Results for the create operations shows "5652.05", but table describing that metric reads "565.05". Both seem very low, but if nothing else they should match.
Question - The argument "tests_str" seems an odd name for this argument. Can this be deprecated in a future release in favor of something like "test_ops"?
BTW, first time I recall reading this material. Very nice, and very helpful!
Attachments
Activity
| Labels | New: easy |
| Assignee | Original: Richard Henwood [ rhenwood ] | New: Lustre Manual Triage [ LM-Triage ] |
| Assignee | Original: Jodi Levi [ jlevi ] | New: Richard Henwood [ rhenwood ] |
| Description |
Original:
25.5 - There are several metrics (test operations) that can be collected by mss-survey. By default four (4) are created: create, setattr, getattr, destroy - in that order. This info should be added to the first section, along with: a) list of minimum operations to run (create and destroy), b) an example of running with the minimum operations c) an example of running with a non-default operation 25.5.1 - In the "Note" about not cleaning up properly, please verify and then add a section about ounting the target type ldiskfs, and recursively deleting the directories under/ROOT/tests*/ if subsequent tests fail to run. I've run across this twice in my limited exposure to this tool. 25.5.2 - The output is truncated. Am not sure what to suggest, as default produces 4 different metrics (long output for each line). Perhaps run with the minimum operations and let it wrap? 25.5.2 - Results for the create operations shows "5652.05", but table describing that metric reads "565.05". Both seem very low, but if nothing else they should match. Question - The argument "tests_str" seems an odd name for this argument. Can this be deprecated in a future release in favor of something like "test_ops"? BTW, first time I recall reading this material. Very nice! |
New:
25.5 - There are several metrics (test operations) that can be collected by mds-survey. In the default execution five (5) are reported: create, lookup, setattr, getattr and destroy - in that order. This "default" info should be added to the first section, along with: a) list of minimum operations to run (create and destroy), b) an example of running with the minimum operations c) an example of running with another supported, but non-default execution (like create, getattr, destroy) 25.5.1 - In the "Note" about not cleaning up files when execution is manually stopped using CNTL-C, please add a section about mounting the target as type ldiskfs, and recursively deleting the "tests*" directories under /mdt/ROOT/ if subsequent executions fail to run. 25.5.2 - The output is truncated. Am not sure what to suggest, as the default produces 5 different metrics (long output for each line). Perhaps run with the minimum operations and let it wrap? 25.5.2 - Results for the create operations shows "5652.05", but table describing that metric reads "565.05". Both seem very low, but if nothing else they should match. Question - The argument "tests_str" seems an odd name for this argument. Can this be deprecated in a future release in favor of something like "test_ops"? BTW, first time I recall reading this material. Very nice, and very helpful! |