Recycling Informix
A DBA (IDS 7.30 on HP-UX 11, 16-CPU OLTP box) asked whether there's any real justification for routinely bouncing an Informix instance, after a consultant urged weekly restarts on 'good practice/gut feeling' grounds. The replies were almost unanimous that there is no technical reason: leave it running unless it crashes, needs a patch/upgrade, or the sysadmins take the machine down, with uptimes of 50-100+ days cited; the only argument for an occasional restart was to prove you can. One poster argued for weekly reboots because of zombie processes, memory leaks and declining process priority, but others countered that those are software bugs to be fixed and that fixed VP priority works fine in 7.31. No formal resolution, but the consensus was against scheduled recycling.
Auto-generated by DrWatson from the posts below — may be imperfect; read the full thread.
Informix Version: 7.30 FC7 OS: HPUX 11.0 Hardware: HP9000 Series 800 V2500 Application: OLTP (strictly database server) Misc. 16 CPU, 8GB mem, 2 instances Folks, Quick question. What is your view regarding recycling the instance. One of our production servers has been up for 30 days. Performance does not appear to be degrading. I say leave it up. Only because I don't have a good argument not to. We have an Informix consultant on site who believes we should bounce every week (good salesman), "because it is a good business practice" "a good gut feeling". (Rudy, you know who I'm talking about?). It may very well be that we should bounce weekly but I would like some hard evidence that this is "... good business practice" Thanks, Regards, Scott Henderson Database Administrator, IS Computing Services ________________________________ Tel: (905) 712-1084 ext. 3740 Fax: (905) 712-6709 e-mail: shenderson@purolator.com ________________________________ Purolator - THIS WAY TO FASTER, EASIER SHIPPING
In article <82m909$buh$1@news.xmission.com>, Henderson, Scott <SHenderson@purolator.com> wrote: >Informix Version: 7.30 FC7 >OS: HPUX 11.0 >Hardware: HP9000 Series 800 V2500 >Application: OLTP (strictly database server) >Misc. 16 CPU, 8GB mem, 2 instances >What is your view regarding recycling the instance. One of our production >servers has been up for 30 days. Performance does not appear to be >degrading. I say leave it up. Only because I don't have a good argument not >to. >We have an Informix consultant on site who believes we should bounce every >week (good salesman), "because it is a good business practice" "a good gut >feeling". (Rudy, you know who I'm talking about?). It may very well be >that we should bounce weekly but I would like some hard evidence that this >is "... good business practice" Scott: Some business practices preclude routine bouncing. We embed an Informix server in one of our products. It's expected to run non-stop, except for scheduled maintenance every few weeks. We've delivered various versions of this system using Informix 7.14, 7.23, and 7.31 on HP-UX 9.05, 10.10 and 10.20, on both uniprocessors and multiprocessors, and have seen no problems with letting the server run until the next patch, upgrade, or application fault. Jim -- W. Jim Jordan, Nortel Networks, Stop 29CA3A08 | +1 613 763 1568 PO Box 3511 Stn C, Ottawa, ON K1Y 4H7 Canada | wjjordan@nortelnetworks.com Opinions expressed are not necessarily those of Nortel Networks. Inbound spam filtering is in place. Don't send what I won't see.
I see no reason to bounce the server periodically from a technical standpoint. The main reason that I can think of to bounce the server is the same reason that I can think of to bounce the server-machine. That is 'to make sure that you can'. Henderson, Scott wrote: > Informix Version: 7.30 FC7 > OS: HPUX 11.0 > Hardware: HP9000 Series 800 V2500 > Application: OLTP (strictly database server) > Misc. 16 CPU, 8GB mem, 2 instances > > Folks, > > Quick question. > > What is your view regarding recycling the instance. One of our production > servers has been up for 30 days. Performance does not appear to be > degrading. I say leave it up. Only because I don't have a good argument not > to. > > We have an Informix consultant on site who believes we should bounce every > week (good salesman), "because it is a good business practice" "a good gut > feeling". (Rudy, you know who I'm talking about?). It may very well be > that we should bounce weekly but I would like some hard evidence that this > is "... good business practice" > > Thanks, > > Regards, > Scott Henderson > Database Administrator, IS Computing Services > ________________________________ > Tel: (905) 712-1084 ext. 3740 > Fax: (905) 712-6709 > e-mail: shenderson@purolator.com > ________________________________ > Purolator - THIS WAY TO FASTER, EASIER SHIPPING
On Wed, 8 Dec 1999 13:21:58 -0500, "Henderson, Scott" <SHenderson@purolator.com> wrote: > >Informix Version: 7.30 FC7 >OS: HPUX 11.0 >Hardware: HP9000 Series 800 V2500 >Application: OLTP (strictly database server) >Misc. 16 CPU, 8GB mem, 2 instances > > >Folks, > >Quick question. > >What is your view regarding recycling the instance. One of our production >servers has been up for 30 days. Performance does not appear to be >degrading. I say leave it up. Only because I don't have a good argument not >to. > >We have an Informix consultant on site who believes we should bounce every >week (good salesman), "because it is a good business practice" "a good gut >feeling". (Rudy, you know who I'm talking about?). It may very well be >that we should bounce weekly but I would like some hard evidence that this >is "... good business practice" Did you ever heard of zombie processes? Did you ever heard of memory lacks? I think it is really a good practice to reboot the server on which the instance runs cause of buggy software that does not release memory and so on. Another reason: the priority of processes decreases the longer a process runs. You can specify fixed priority for your VP's but the last I know about it is that it is buggy. I tried it with Online 7.14 and there where running only the VP's and nothing else .... That was with Reliant Unix 5.43 on Siemens RM600. Once a week reboot is a dood idea... only my personal opinion. Reinhard
I say leave well enough alone. My engines only bounce when they crash or when the system is booted by the sysadmins for maintenance. Period. Art S. Kagel "Henderson, Scott" wrote: > > Informix Version: 7.30 FC7 > OS: HPUX 11.0 > Hardware: HP9000 Series 800 V2500 > Application: OLTP (strictly database server) > Misc. 16 CPU, 8GB mem, 2 instances > > Folks, > > Quick question. > > What is your view regarding recycling the instance. One of our production > servers has been up for 30 days. Performance does not appear to be > degrading. I say leave it up. Only because I don't have a good argument not > to. > > We have an Informix consultant on site who believes we should bounce every > week (good salesman), "because it is a good business practice" "a good gut > feeling". (Rudy, you know who I'm talking about?). It may very well be > that we should bounce weekly but I would like some hard evidence that this > is "... good business practice" > > Thanks, > > Regards, > Scott Henderson > Database Administrator, IS Computing Services > ________________________________ > Tel: (905) 712-1084 ext. 3740 > Fax: (905) 712-6709 > e-mail: shenderson@purolator.com > ________________________________ > Purolator - THIS WAY TO FASTER, EASIER SHIPPING
I'm running a 24x7 mixed OLTP/DSS DB on SCO. The last time I rebooted one of my two servers (one always stays up as a hot backup) - and restarted the instance, Informix was up over 100 days: [INFORMIX-OnLine Version 7.22.UC3 -- On-Line -- Up 103 days 14:16:59]. My second server is has been up (initial startup) half that time: [Informix Dynamic Server Version 7.30.UC2 -- On-Line -- Up 56 days 21:57:16]. I typically only shut down the instance when the SA needs to do some maintenance to the server. I say if it works like its suppose to you should never have to restart. Picking an arbitrary date in the future to recycle an instance (1 day, 1 week, 1 month, 1 year) is fine if it makes you feel better - but I'd be more concerned about knowing my data is backed up and the engine is performing well. Just my $0.02
In article <384ed43f.6082736@news.cis.dfn.de>, Reinhard Habichtsberg <Reinhard.Habichtsberg@unilux.de> writes >>What is your view regarding recycling the instance. One of our production > >Did you ever heard of zombie processes? A bug where the parent does not wait for the child. Buggy software needs fixing. >Did you ever heard of memory lacks? Again,Buggy software needs fixing. >I think it is really a good practice to reboot the server on which the >instance runs cause of buggy software that does not release memory and >so on. > I think it's really good practice to get buggy software fixed.. >Another reason: the priority of processes decreases the longer a >process runs. You can specify fixed priority for your VP´s but the >last I know about it is that it is buggy. I tried it with Online 7.14 >and there where running only the VP´s and nothing else .... That was >with Reliant Unix 5.43 on Siemens RM600. > It works fine on 7.31. >Once a week reboot is a dood idea... > >only my personal opinion. > >Reinhard -- David Williams