Description
Increasing customer reports of storage alarms with “poll error” on Linux systems. This is a legitimate indication that core’s memory usage is high enough to prevent us from executing storage monitoring utilities, and user should somehow reduce memory usage.
<br>
However, server 9.9.8 introduces a redesign of SysmgmtPI that eliminates the occurrence of poll errors, by moving utility execution into the exacqd process, which remains at low memory usage. In turn, storage alarms will be more meaningful to users, and this will also reduce the frequency of all-encompassing storage alarms that are essentially false positives.
<br>
Platforms
Linux 16.04 x64 (any Linux is vulnerable, but Windows is safe)
<br>
Steps to reproduce
Keep adding streams to increase core’s memory usage. Can probably exacerbate with free run recording, higher resolution/quality, etc.
<br>
Expected result
Drives remain healthy, and various storage hardware operations like self-test or driveprep should remain functional.
<br>
Actual result
All drives become alarmed with “poll error”.
<br>
Work around
Reduce memory usage (less streams, lower resolution, restrict recording) to keep core’s usage at below 50% of physical memory.
<br>
Resolution
exacqServer 9.9.8 or later
<br>
How-to-eliminate-storage-alarms-with-poll-error.pdf