Server Crash

> -----Ursprungligt meddelande-----
> Från:	Paul Andrews 
> Skickat:	den 8 november 1999 10:14
> Ämne:	Ingres II server crash
> 
> We've been running Ingres II on HP-UX 11.0 for a few weeks now and generally
> things have now settled down and are going well. So far we are averaging a
> server crash (without trace) every week (two in the last week). We are
> running two DBMS servers on a four processor machine. Anyone else seen this?
> 
> Paul

From: Conny Gyllander 
Subject: SV: Ingres II server crash

We had a problem, that occurred approximately once a week and after a while we
found out what was wrong.

If you have the environment variable II_GC_PROT = UNIX, then you have the
sockets sitting in /tmp by default. 

If you then have a script that automatically removes everything under /tmp
that is older than lets say 5 days, then your instance comes to a nasty halt
because the sockets will also go when they are old enough.
To fix this: as root 
 # mkdir /ING; chmod 755 /ING; chown ingres /ING 
 and as ingres 
  $ ingsetenv II_GC_PORT /ING/ 

The new directory to hold the socket files can be any name, but if you 
stick to 3 letters in length, then you can still see servers in IPM.
Don't forget the / at the end  of the II_GC_PORT value. 

//Conny
Ingres Q & A
Back to William's Home Page
Email William