Hanging CPU VP

Problem: Hanging CPU VP Hardware: Fujitsu M754i 4 x Pentium Pro 200 1.5Gb RAM 22 x 4Gb Hard Drives (11 For Mirror) OS: UnixWare 2.1.3 (spit spit spit) Engine: IDS 7.23UC1 I'm after a deeper understanding of our current problem. The problem is that on occassion 1 of our 3 CPU VPs hangs (truss -p returns nothing but a flashing cursor!). Informix's advise is to upgrade our engine - unfortunately the next release requires UnixWare 7 and that would mean we'd have to migrate .. this option is not currently viable :(. A decision was made that we would 'live' with this problem until the hardware reached the end of its life cyle. However, of late the problem has become more frequent (we suspect due to development work) and the downtime is becomming unacceptable. I would appreciate any thoughts/experience you have this regards to this problem. Thanks in advance Andrew Bland DBA City of Bradford Metropolitan District Council

Andrew Bland <blanda@hep.bradford.gov.uk> wrote in message news:3832B418.DCF0A698@hep.bradford.gov.uk... > Problem: Hanging CPU VP > Hardware: Fujitsu M754i 4 x Pentium Pro 200 1.5Gb RAM 22 x 4Gb Hard > Drives (11 For Mirror) > OS: UnixWare 2.1.3 (spit spit spit) > Engine: IDS 7.23UC1 > > I'm after a deeper understanding of our current problem. The problem is > that on occassion 1 of our 3 CPU VPs hangs (truss -p returns nothing but > a flashing cursor!). Try to turn off multiprocessing. I mean set MULTIPROCESSOR=0 SINGLE_CPU_VP = 1NUMCPUVPS=1 What about messages in your log file ? Also, you must found af.###### file with system dump. Send this file too. HTH. ------------------------------------------------- With best regards, Yuri Dovgart SAP R/3, Informix technical consultant, Informix Certified Professional, Senior System Consultant System Architecture and High Availability Systems, 'Telecominvest' company Email y_dovgart@tci.ukrtel.net ICQ 39284285

It's going to be difficult to figure out what the engine is doing by using truss because the CPUVPs make very few system calls. This is especially true if you are using NETVPs and AIOVPs. Probably the best thing to do is to check to see what is running on the CPUVPs when it gets hung. Run onstat -g ath several times. If the cpuvp that is hung is always running the same sqlexec thread, then it's possible that you have a hung thread. From that you should be able to figure out which session is associated with the thread. Then you can run onstat -g ses and figure out what the query is that is running when the problem occurs. Once you have the query, you will need to ensure that you can force the problem by manually running the query in dbaccess. If you can force the problem, then you need to get a stack trace of the CPUVP when the problem is occuring. On some platforms this can be done by doing a kill -7 to the CPUVP that is hung. I don't know for sure if this will work on your system. If it in the optimizer you'd see sqoptim() in the stack trace. If you see that and if the query causing the hang containg a join between two tables, then check to see if the query is joining an int to a char field. If that is the case, then you might be able to work arround the problem by dropping distributions on one of the two columns. I've only seen one situation where the CPUVP would hang like this and it involved such a join. I can't rember when it was fixed, but it was fixed several years ago and - well - 7.23UC1 is several years old. Andrew Bland wrote: > Problem: Hanging CPU VP > Hardware: Fujitsu M754i 4 x Pentium Pro 200 1.5Gb RAM 22 x 4Gb Hard > Drives (11 For Mirror) > OS: UnixWare 2.1.3 (spit spit spit) > Engine: IDS 7.23UC1 > > I'm after a deeper understanding of our current problem. The problem is > that on occassion 1 of our 3 CPU VPs hangs (truss -p returns nothing but > a flashing cursor!). Informix's advise is to upgrade our engine - > unfortunately the next release requires UnixWare 7 and that would mean > we'd have to migrate .. this option is not currently viable :(. > > A decision was made that we would 'live' with this problem until the > hardware reached the end of its life cyle. However, of late the problem > has become more frequent (we suspect due to development work) and the > downtime is becomming unacceptable. I would appreciate any > thoughts/experience you have this regards to this problem. > > Thanks in advance > > Andrew Bland > DBA > City of Bradford Metropolitan District Council

I can't go down to a single cpu because the system would run like a dog, I have 300 users concurrently accessing the database. Sometime the hung processor can be brought back to life by going to super user and issuing a wall message! This occured this morning. However, this doesn't always 'fix' the problem. Hmm if we manage to 'fix' the problem the informix log fills up with this message listener-thread: err = -25587: Network receive failed. all users connect via IPCSHM so this message seems to indicate that the poll thread can't process ipcshm requests. But I'd like to thank everyone who has replied so far, all the ideas are been taken into consideration .. even the wise crack about unixware's comparison to a spreadsheet/calculator (we are aware of this problem). Yuri Dovgart wrote: > Andrew Bland <blanda@hep.bradford.gov.uk> wrote in message > news:3832B418.DCF0A698@hep.bradford.gov.uk... > > Problem: Hanging CPU VP > > Hardware: Fujitsu M754i 4 x Pentium Pro 200 1.5Gb RAM 22 x 4Gb Hard > > Drives (11 For Mirror) > > OS: UnixWare 2.1.3 (spit spit spit) > > Engine: IDS 7.23UC1 > > > > I'm after a deeper understanding of our current problem. The problem is > > that on occassion 1 of our 3 CPU VPs hangs (truss -p returns nothing but > > a flashing cursor!). > > Try to turn off multiprocessing. I mean set > > MULTIPROCESSOR=0 > SINGLE_CPU_VP = 1> NUMCPUVPS=1 > > What about messages in your log file ? Also, you must found af.###### file > with system dump. Send this file too. > > HTH. > > ------------------------------------------------- > With best regards, Yuri Dovgart > SAP R/3, Informix technical consultant, > Informix Certified Professional, > Senior System Consultant > System Architecture and High Availability Systems, > 'Telecominvest' company > Email y_dovgart@tci.ukrtel.net > ICQ 39284285