RE: Extremly long checkpoints: How to find the reason and solve t

Joerg, If You are using FC, it can be a VERY low-level hardware issue: When FC laser is dying, the disk response time becomes very big, especially under load. I saw that problem in the past with HP V-class cluster. The temporary solution was to switch off one of cluster nodes. The final solution was to replace one of FC controllers. Everything works fine since the controller was replaced. ------------------------------------------ Alexey Sonkin > -----Original Message----- > From: js@jetsys.de [mailto:js@jetsys.de] > Sent: Tuesday, May 18, 2004 1:24 AM > To: informix-list@iiug.org > Subject: Re: Extremly long checkpoints: How to find the reason and solve > the problem (7.31 FD3) > > roeferr@ig.com.br (Ferronato) wrote in message > news:<df1bdeb4.0405131055.4c74eeef@posting.google.com>... > > Hello, > > > You didn't mention , but I can guess that you are using a box HP, with > > 64 bits HP-UX 10.xx or like, write ? Confirm it,please. > > I'm not sure the version of IDS , but I know that in a company where I > > were working 2 years ago, we had the same problem. The solution was > > update the OS HP-UX and the database version. There was bugs in both > > (OS and IDS). > > machine is about 5 years old and a DEC Alpha cluster with True Unix > 4.0f. We finally found the reason for the long checkpoints. Some kind > of hardware problems with the mirrors. HW configuration are lots of > SCSI drives connected to a fibre channel controller connected to a > fibre channel switch connected with the HBA on the cluster. With > active mirrors we get the long checkpoints. With disabled mirrors > everything is fine. First idea from hp support was to low level > reformat the mirror disks and rebuild the mirror. Low level format > succeeded without problems but after rebuilding the mirror, the > problem reoccurs. So it's possibly one of the others components. > > Greetings, Joerg sending to informix-list