To some extent these may or may not be "gotcha's". Background. IBM z/890 with 1 IFL with FICON attached dasd. IFL has 4 GB memory. Running Oracle 10g with OEM. Also a variety of NFS servers, FTP servers, Samba servers, VSE Virtual Tape servers, DB2/UDB test machines, Web servers, DB2 Connect servers and some basic Linux machines. Most of the servers are underutilized/idle. We have 4 Oracle servers, a production one, a test one (same as production with data refreshed from production), development one (table definitions may change compared to the production one) and my DBA one (where I test things out before I impact the other images).
Based on the documentation from Oracle, in my case of Suse 10 SP 2 with Oracle 10g R2, there may be some RPMs that are needed and changes to some parms (such as sysctl.conf, Limits.conf, login, profile.local) are needed. It is much easier if your first Oracle installs mirror the IBM Redbooks. Of course that won't be the latest Oracle, nor the latest Linux, but then you see the process and the result. If you train by using the latest Linux with the latest Oracle, you will have months of fun <G>. The production server has 300 users defined, and normally 95 active during 1st shift. The database size is 7 GB. The production database normally runs 20-25% of the z/890 IFL. OEM (Oracle Enterprise Manager). Like most products, it doesn't understand virtualization. The maximum CPU line on its graphs are bogus. At times, when some batch type process is running, it show we are using 3.5 to 4 CPUs. Instance disk I/Os shows past 40,000 per second vs my monitor shows perhaps 500 per second. I use the graphs to give me a relative idea of what is happening. When another machine drives the IFL to 100%, OEM shows we are having problems. However, with the CP share of 1000 vs 100 for others, there is no real contention and users don't have any complaints. Ooops. Almost forgot. I trade memory savings for a higher I/O rate. This image is 750 MB. (The other Oracle images are 600 MB). To get the image size down that small, I had to make the SGA size about the smallest it can be (with OEM running) of 200MB, (140 MB on the other images). The gotcha, is when you install Oracle, I think it required a 1 GB machine to install, it started to set memory parms. You have to go back and adjust them for your size. And then I had to change the PGA size also. If you really need performance, then the trade is you loose easy administration. RAW file access against native dasd spread across controllers. Pain if you need more space. I didn't need that type of performance so my tablespaces are in a LVM pool. Need more space, add a pack to the LVM pool. No changes to Oracle. Oracle assumes that you are on crap hardware. It wants to put copies of its log files on separate disks. I had a small fight with it about that. We didn't have FCP tape drives attached, so I do two types of backups. 1. I cycle Oracle and do a flash copy each week. I DDR the images to escon 3590s for disaster backups. 2. A copy of Oracle is exported and the logs are copied to an NFS server. The files on the NFS server are tar'ed and put on a 3590 tape. (I can reload via tar any/all files back to the NFS server.) The documentation, even when discussing zSeries processors, have requirements that are PC in nature. i.e. cheap CPU, cheap memory, crap I/O. However, if you are seriously going to drive a couple IFL engines, they are good starting points. Of course, not having a good VM performance monitor is a serious gotcha if you are doing heavy loads or where performance is critical. We had a developer issue a query from hell which tied up the server for hours, but Oracle continued to service the other queries in a timely fashion. Not hard when you are only using 25% of the processor, normally. You might need a startup/shutdown script to activate/deactivate lsnrctl, the database, and dbconsole. That's all I can thing of for now. We have been running for 4 years now. No failures. No problems. No maintenance added. Just runs. But then, never had a disaster recovery attempt. Never had to reload a single table from backup. I have restored the database from backup but didn't need to roll forward. Tom Duerbusch THD Consulting >>> "Davey, Clair" <cda...@scspa.com> 2/15/2011 10:25 AM >>> I would like to hear some of your 'gotcha's when running Oracle databases on Linux on Zeries. You can send them offline if you would like. The contents of this e-mail are confidential to the ordinary user of the e-mail address to which it was addressed and may also be privileged. If you are not the addressee of this e-mail you should not copy, forward, disclose or otherwise use it or any part of it in any form whatsoever. If you have received this e-mail in error, please notify us by telephone or e-mail the sender by replying to this message, and then delete this e-mail and other copies of it from your computer system. We reserve the right to monitor all e-mail communications through our network. ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For more information on Linux on System z, visit http://wiki.linuxvm.org/ ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For more information on Linux on System z, visit http://wiki.linuxvm.org/
