[LU-1127] need performance-sanity to work on Hyperion Created: 21/Feb/12 Updated: 15/Dec/14 Resolved: 15/Dec/14 |
|
| Status: | Resolved |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | Lustre 2.2.0, Lustre 2.3.0, Lustre 2.1.1 |
| Fix Version/s: | None |
| Type: | Bug | Priority: | Minor |
| Reporter: | Cliff White (Inactive) | Assignee: | Minh Diep |
| Resolution: | Won't Fix | Votes: | 0 |
| Labels: | None | ||
| Environment: |
Hyperion |
||
| Severity: | 3 |
| Rank (Obsolete): | 10106 |
| Description |
|
To have proper repeatable performance testing on hyperion, our performance-sanity test needs to work. Performance-sanity can then be used for the performance section of release testing. |
| Comments |
| Comment by Andreas Dilger [ 03/May/12 ] |
|
Is there a particular reason that we are using mdsrate in preference to mdtest? I understand that mdsrate was written originally for Lustre metadata testing, but I wonder if we could move over to mdtest and get equivalent functionality? My question/concern is about using the more standard mdtest benchmarking tool for Lustre performance testing, so that it is easier to compare results from mdtest to mdsrate. My comments aren't necessarily a request to change the tool being used right now, but more to start a discussion of whether we should move away from mdsrate in our scale testing. |
| Comment by Minh Diep [ 03/May/12 ] |
|
Majority of the time we had been using mdtest on hyperion. AFAIK, there are a few time that we needed to use mdsrate instead of mdtest were when mdtest hit a blocker while mdsrate did not. I noticed that DNE project were using mdsrate. Wang Di modified mdsrate to run with multiple mounts. Perhaps we should find out if we can do the same with mdtest. Adding Di to this. |
| Comment by Cliff White (Inactive) [ 03/May/12 ] |
|
We've also had some issues keeping mdsrate running at Hyperion, I would not object to dropping it. As Minh said, we routinely run mdtest. |