Operating System - HP-UX
1748250 Members
3445 Online
108760 Solutions
New Discussion юеВ

IO Timeout on an informix db.

 
someone_4
Honored Contributor

IO Timeout on an informix db.

Hello
Right now we are having an issue where we see the attachted errors on our informix database server. It is an N4000 conected over fiber to an XP-256.

Now I have already determined that the timeout is because the main device is not responding in the about of time set in the IO Timeout for that device. The errors were pin pointed to two devices and i increated the IO timeout to 180 from the default of 30. But I still have some questions that I need help on.

1. The database performance has decreased. And increasing the timeout value did not help. When I mean performance I mean we can not run updates on the database at all. The issue started last night this morning. And yesterday we were able to do updates. How can I go about trouble shooting this issue? Would it be related or do we have two different issues?

2. How can I go about trying to find out why the device was slow to respond? What I am looking for is tips on digging up the root cause of the timeouts on the two devices.

Thanks
Richard
2 REPLIES 2
Eugeny Brychkov
Honored Contributor

Re: IO Timeout on an informix db.

Richard,
can you tell for sure that other LUNs then indicated do not experience such issues? I mean is it whole XP512 behavior or just two LUNs? Please note that timeouts can occur only on LUNs being accessed. You will never see timeout for device not being accessed.
Some more questions:
- how your SAN is configured: fabric, loop? If fabric, then analyze switch diag output (for brocade it's 'supportshow'), if loop then check tdutil stat output; if there's misbehaving loop device then you should find it simply removing devices one-by-one (or adding previously removed ones);
- I see that there're alternate paths configured for XP. Can you check if this faulty path is a primary link for other failing LUNs? Again, if it's alternate path, then system will not use it because another (primary) XP path is healty.
Failed path is c9t3d5
Alternate c10t3d5
So make sure c9 is healty - check if other failing LUNs have c9 as a primry path, then simply troubleshoot this c9 and everything connected to it.
We will be able to help you if you'll provide SAN connectivity description and switch supportshow (if available)
Eugeny
Zafar A. Mohammed_1
Trusted Contributor

Re: IO Timeout on an informix db.

If the devices are connected to EMC box, I know there is one bundled package from EMC where you can check the DISK health and other information also. Maybe you already have contract with EMC or HP, try to investigate immediately with those guys.

Thanks
Zafar