Bug #3910 StopOnError without effect
Submitted: 27 May 2004 9:41 Modified: 8 Jun 2004 13:08
Reporter: Johan Andersson Email Updates:
Status: Closed Impact on me:
None 
Category:MySQL Cluster: Cluster (NDB) storage engine Severity:S2 (Serious)
Version:MySQL 4.1-alpha bk OS:Linux (RH 9)
Assigned to: Jonas Oreland CPU Architecture:Any

[27 May 2004 9:41] Johan Andersson
Description:
I have noticed (running ndbcluster interactively) that if a db node shuts down because of an error, then it will automatically restart, even though StopOnError is set to Yes. This can lead to wild things especially during system (re)start.

Found this problem/feature in MySQL 4.1-alpha bk, on Linux 2.4.2x (RH 9).

How to repeat:
Start a 2-db cluster (2 replicas) interactively
Ctrl-c one node -> Yoyo.
[27 May 2004 13:40] Magnus BlÄudd
I have investigated a little. One problem is that a SIGSEGV in the angel when it calling waitpid. And the child performs abort(). I tried ignoring SIGABRT in combination with the signal "recevied" originally.
Then I got two core files, one that really was wanted, and the other one for the SIGSEGV in waitpid.
This was VM_TRACE'ed of course.

But I imagine that if a SIGSEGV or similar happens in the angel it will restart over and over. Isn't the data that theConfig is pointing to, deleted by the child??
[28 May 2004 9:10] Jonas Oreland
I tried with (in debug mode) both
cntr-c, watch dog crash(error 9998) and ndbrequire(9997) 

But it worked fine.
I did however find a bug in the code, which I fixed.
[8 Jun 2004 12:56] Johan Andersson
Ok!