in the past couple of months, we experience QVS crashes with the log message "300 Error Restart: Server aborted trying to recover by restart. Reason for restart: No document purger activity for 120 seconds".
On bad days, it may happen 10 to 15 times. Sometimes the crashes are 7 minutes apart, up to 4 consecutive times.
Naturally, this causes a very bad user experience.
We optimized our document and monitored the server's memory, CPU and IO consumption - nothing out of the ordinary, plenty of resources available.
We tried to pinpoint with detailed audit log what's the trigger - found no clue yet.
Has anybody encountered this issue?
Can "120 seconds" be tuned - perhaps a larger value is required in our case?
Solved! Go to Solution.
A happy update:
Following troubleshooting sessions with Qlik, the problem was pinpointed to the layout-state-saving mechanism.
In our major document we have over 3,000 objects (). QV automatically saves and loads their layout state every reload.
When it takes longer that 2 minutes.... The Phoenix watchdog kicks in.
A special test build that allows disabling of this mechanism was created by Qlik and tested by us - seems to solve the problem and greatly improve the reload experience.
Since our users don't modify the layout anyway, this solution suits us.
Stay tuned for SR13 - we were told it is incorporated there in a parameter called EnableApplyLayoutState in Settings.ini, enabled by default (compatible with existing behavior), but can be set to 0.
When the purger thread gets hanged, and subsequently killed by QVS self-preserving safety mechanisms by a restart, the message does not indicate any cause.
The issue is not possible to debug either, since when the error occurs (=phoenix restart), the thread is already killed and the call stack gone.
I would check the .share files to make sure there are no issues there.
about 48 hours ago we cleaned the .shared file, it had a few "OrphanEntry" objects.
Since then, things have improved considerably.
We'll follow up in a few more days.
Our hopes were high, but as of 8 hours ago QVS started crashing exactly every 5 minutes by Phoenix.
We upgraded to the latest SR, no change.
We even removed all the content folders - crashing continues....
Has anyone encountered anything like it?
Try turning off the performance logging in the QMC for the Qlik server.
Disabled performance logging, even restarted services - still 5 minutes (to be precise: 5:20~5:30 between restarts)
I don't have a cure but...
I faced something similar a few weeks ago, but in my case revoming of .shared files still helps. Thinking about session collaboration, maybe it's worth to turn off it.
Also I found that "Add Bookmark" web dialog (shown by QVWS, e.g. via Access Point) does not work, it just has not OK button anymore. And this concerns session collaboration feature as well as .shared files. QVS release is 11.20.12758.0.
Found and tracking several posts connected with all of this:
This may helps.
This thread may makes you to think in different way to solve this issue, not sure.
We tried - without success - the following:
Today we simply set up a new server.
I hope we'll get someday to checking the performance counter rebuilding tip.
Hi everyone, thanks for the advice so far.
Today we encountered another restart, this time with the following reason:
2015-06-08 07:34:45 2015-06-08 12:40:25 1 300 Error Restart: Server aborted trying to recover by restart. Reason for restart: Phoenix detected: Performance monitor stuck
The server is new (fresh Amazon Windows 2012R2 instance). Any idea what could "corrupt" the performance counters, as some suggested?