Symptômes
Changes in dce.pl and getPerformance.pl scripts to avoid gaps in CloudIQ graphs for Unity arrays.
Cause
getPerformance.pl is the script responsible for generating the performance metrics.
dce.pl does these steps:
1. Call getPerformance.pl
2. Zip up the performance file gathered
3. MFT the file
4. Remove the job directory
5. Release the lock
6. Exits
dce.pl lock makes sure one job/task to collect the performance metrics is running at a time for same job, and at most one same job waiting in the queue.
The getPerformance.pl is started in dce.pl for cloudOpMgmt_perfMetrics_hifreq_v1 and gets the lock (1st job), so when getPerformance.pl sleep more, the next dce.pl for the same job may wait for more time on the lock (2nd job which will wait in the queue), and the third dce.pl for the same job will be skipped if there's already one waiting on the lock when the third dce.pl is started.
Also, in dce.pl, we check for the lock once in 30 seconds before allowing a performance cycle to start. This will add to the total time the process takes to complete.
Résolution
getPerformance.pl and dce.pl scripts have been modified to collect the performance metrics efficiently and transfer them to CloudIQ. The fix will be available in 5.1 release.
- The getPerformance.pl changes apply to 5.0.x and older but not to 5.1. getPerformance.pl changes will not be used in V5.1, since the method of obtaining performance files is changing in v5.1.
- The dce.pl changes apply to all releases, including 5.1.
Change in getPerformance.pl script:
waitTime that the getPerformance.pl waits before requesting the performance data has been changed.
Change in dce.pl script:
1. dce.pl script has been modified to check for the lock every 10 seconds. This allows the script to start earlier and reduces the total time that the script runs.
2. In previous codes, MFT operation to transfer the performance data was performed first and then release the lock, increasing the total time that the process takes. In the modified script the LOCK will be released before the MFT transfer operation, allowing the next performance cycle to start earlier.
Workaround:
Modified scripts have been attached to this KB.
Dell EMC Support team will assist in applying the change.
Informations supplémentaires
IMPORTANT NOTE:
===============
Please confirm if the issue is the same before applying the changes.
If/when the array is updated from current release to a newer release that does NOT have the fix, the fix will need to be manually applied again. Otherwise, gaps might reappear. Dell EMC will be able to reapply the fix as needed.
23 nov. 2020