Hi, Question#1:
We have a F740 filer used as the Oracle DB storage. Our platform is : Sun-Solaris 2.7, SUN 420R,using Giga-bit private connection into filer; using recommended NFS mount options for Oracle 816. Situation: when we have a request to "expand" our filer for sharing (exportfs), and as the time the Oracle daemons were running. We changed the filer's /etc/exports, then executed "exportfs -au", "exportfs -a".
And, right after those 2 command to the filer, ALL our Oracle daemons crashed and died. Is that supposed to happen?
Question#2:
We tried to write a SHELL script to automate our daily (filer) snapshot operation. During the course of testing, we found out a very annoying situation. (The HW/SW specification are the same as above.)
There is a Q-tree called "/vol/vol01/ismis_oracle_u1" and is mounted under this machine at the mount point of "/oracle/u1". The problem happened after the snapshot was taken.
One of our team member set the source PATH (wrong) to /oracle/u1/oradata/.snapshot in the script (supposed to be "/oracle/u1/.snapshot"). And, this PATH actually worked some of the time. But, when it failed , it showed the error message as "cp: can not find files /oracle/u1/oradata/.snapshot/*". Interesting thing is if we insert a "/bin/ls /orcale/u1/oradata/.snapshot/*" right before the "cp" command, the directory seems to be there after the command , hence the "cp" operation was happy.
Any idea? Thanks!
Jack Li UNIX System Admin. Team, Toronto Production Support, Capital Markets, Nesbitt Burns Emf!sys, Division of Bank of Montreal Group of Company 416-867-3527,(fax)416-867-7157 mailto:Jack.Li@bmo.com Jack.Li@bmo.com
On 7 Sep 2000, at 10:13, Li, Jack wrote:
Hi, Question#1: Situation: when we have a request to "expand" our filer for sharing (exportfs), and as the time the Oracle daemons were running. We changed the filer's /etc/exports, then executed "exportfs -au", "exportfs -a".
And, right after those 2 command to the filer, ALL
our Oracle daemons crashed and died. Is that supposed to happen?
I've had a similar problem. My filers are F740's with dedicated gigabig connections to IBM RS/6k-F50 processors. It works quite well, but have had several problems.
a) Similar to yours . . . . if I add/change a export, at times AIX will start having problems accessing the db files. The mount points seem to be fine - I can cd to the filesystems, vi files, copy files, etc, but Oracle seems to have some kind of problem.
b) With each of my 2 files, at separate times, I've had Oracle log problems. For some reason Oracle can't write to one of the log files. Everything seemed fine - I could cd to the filesystems, copy files, create files, etc, but Oracle could kept getting errors writing to one of the log files - even after deleting and recreating the log file!!!!! Rebooting the oracle server didn't help!!!! Finally, rebooting the filer fixed the situation. After this happened to the first oracleserver/filer I figured it was a fluke, but then it happened to my other oracleserver/filer system.
Rick
Question#2: We tried to write a SHELL script to automate our daily (filer)
snapshot operation. During the course of testing, we found out a very annoying situation. (The HW/SW specification are the same as above.)
There is a Q-tree called "/vol/vol01/ismis_oracle_u1" and is mounted
under this machine at the mount point of "/oracle/u1". The problem happened after the snapshot was taken.
One of our team member set the source PATH (wrong) to
/oracle/u1/oradata/.snapshot in the script (supposed to be "/oracle/u1/.snapshot"). And, this PATH actually worked some of the time. But, when it failed , it showed the error message as "cp: can not find files /oracle/u1/oradata/.snapshot/*". Interesting thing is if we insert a "/bin/ls /orcale/u1/oradata/.snapshot/*" right before the "cp" command, the directory seems to be there after the command , hence the "cp" operation was happy.
Any idea? Thanks!
Jack Li UNIX System Admin. Team, Toronto Production Support, Capital Markets, Nesbitt Burns Emf!sys, Division of Bank of Montreal Group of Company 416-867-3527,(fax)416-867-7157 mailto:Jack.Li@bmo.com Jack.Li@bmo.com
Jack.Li@BMO.com (Jack Li) writes [...]
Situation: when we have a request to "expand" our filer for sharing
(exportfs), and as the time the Oracle daemons were running. We changed the filer's /etc/exports, then executed "exportfs -au", "exportfs -a".
And, right after those 2 command to the filer, ALL
our Oracle daemons crashed and died. Is that supposed to happen?
I would certainly expect it to. You've got a window when the filing systems are not exported to the Oracle servers. If they read or write their files in that window they will get an NFS access error from the filer, and will very likely be unable to recover.
Why are you using "exportfs -au" at all? "exportfs -a" by itself will update the exports for everything mentioned in /etc/exports. It doesn't deal with the case when you've *removed* a whole entry from /etc/exports, but then neither does your procedure.
Chris Thompson University of Cambridge Computing Service, Email: cet1@ucs.cam.ac.uk New Museums Site, Cambridge CB2 3QG, Phone: +44 1223 334715 United Kingdom.