LOGSIZE recommendation
Hi. I'm currently preparing to migrate our Informix databases to new servers in the datacenter of our parent company. Also configuring an HDR pair for high availability. We've run some performance tests, including an application mass update. When this test, and the nightly update cycle run, the logs offload about once every 1-2 minutes. This seems a bit excessive, however, I realize there is a tradeoff in checkpoints/recovery time/data loss/sync'ing HPL, etc. So I thought I'd ask for some recommendations from this group. Current specifications are LOGSIZE=2000 LOGFILES=10 LOGBUFF=32 Logs are offloaded via the alarmprogram.sh using ONBAR and Netbackup to a virtual tape library, pretty quick. Note that this database/application are from a 3rd party vendor who supplied most of the database configuration parameters. Thanks for your assistance
I would still increase the log size and the number of logs for replication and transaction processing. What you are showing seems low to me. You need to determine the amount of work that it will take to fill up a number of logs in a critical period or day. Thanks, ************************************** Ernie Knox Sears Holding Co. IT Database Administrator Specialist IT Service Management, Strategy & Architecture 3333 Beverly Rd., B4-266A Hoffman Estates, IL. 60179 Office: (847) 286-5735 Fax: (847) 645-3874 Pager: (800) 759-8352 Pin#: 7271042 Email: eknox@sears.com " It's always a great day to watch Football ! " ************************************** -----Original Message----- From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of GARY PRICE Sent: Thursday, October 18, 2007 10:51 AM To: ids@iiug.org Subject: LOGSIZE recommendation [10196] Hi. I'm currently preparing to migrate our Informix databases to new servers in the datacenter of our parent company. Also configuring an HDR pair for high availability. We've run some performance tests, including an application mass update. When this test, and the nightly update cycle run, the logs offload about once every 1-2 minutes. This seems a bit excessive, however, I realize there is a tradeoff in checkpoints/recovery time/data loss/sync'ing HPL, etc. So I thought I'd ask for some recommendations from this group. Current specifications are LOGSIZE=2000 LOGFILES=10 LOGBUFF=32 Logs are offloaded via the alarmprogram.sh using ONBAR and Netbackup to a virtual tape library, pretty quick. Note that this database/application are from a 3rd party vendor who supplied most of the database configuration parameters. Thanks for your assistance ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum.
I manage an IDS 10 db behind a 3rd party app and have made numerous changes to accommodate performance and growth. The vendor had no problem with us doing it. Shouldn't be any reason you couldn't: - add more TX logs - add more larger TX logs - change LTAPEBLK for faster throughput - maybe play with LOGBUFF OS version? IDS version? Bob Roussey Unix / Informix Administration Spirit Airlines Robert.Roussey@SpiritAir.com -----Original Message----- From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of GARY PRICE Sent: Thursday, October 18, 2007 11:51 AM To: ids@iiug.org Subject: LOGSIZE recommendation [10196] Hi. I'm currently preparing to migrate our Informix databases to new servers in the datacenter of our parent company. Also configuring an HDR pair for high availability. We've run some performance tests, including an application mass update. When this test, and the nightly update cycle run, the logs offload about once every 1-2 minutes. This seems a bit excessive, however, I realize there is a tradeoff in checkpoints/recovery time/data loss/sync'ing HPL, etc. So I thought I'd ask for some recommendations from this group. Current specifications are LOGSIZE=2000 LOGFILES=10 LOGBUFF=32 Logs are offloaded via the alarmprogram.sh using ONBAR and Netbackup to a virtual tape library, pretty quick. Note that this database/application are from a 3rd party vendor who supplied most of the database configuration parameters. Thanks for your assistance ************************************************************************ ******* Forum Note: Use "Reply" to post a response in the discussion forum.
I agree that the size and number of logs need to be adjusted. Sizing is a trade off between safety on one hand and performance on the other. The smaller the log files the more frequently they will be archived (assuming that you are using continuous log backup or backing up logs using the ALARMPROGRAM as they fill). The larger the log files the less impact on the system. The number of logs depends on three factors: - The size of each logfile - The size of an acceptable large/long transaction - How much recovery data you will need to have on disk ready to recover You should have enough logs (times log size) to allow a valid large transaction plus the expected other update volume that may happen in other sessions during that large transaction plus about 25 - 40%. Art S. Kagel ----- Original Message ----- From: Gary Price <ids@iiug.org> To: ids@iiug.org At: 10/18 11:52:01 Hi. I'm currently preparing to migrate our Informix databases to new servers in the datacenter of our parent company. Also configuring an HDR pair for high availability. We've run some performance tests, including an application mass update. When this test, and the nightly update cycle run, the logs offload about once every 1-2 minutes. This seems a bit excessive, however, I realize there is a tradeoff in checkpoints/recovery time/data loss/sync'ing HPL, etc. So I thought I'd ask for some recommendations from this group. Current specifications are LOGSIZE=2000 LOGFILES=10 LOGBUFF=32 Logs are offloaded via the alarmprogram.sh using ONBAR and Netbackup to a virtual tape library, pretty quick. Note that this database/application are from a 3rd party vendor who supplied most of the database configuration parameters. Thanks for your assistance ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum.
Your current specifications are very low LOGSIZE=2000 LOGFILES=10 Your total log space is 2000*10=20000=~20MB (we have ~7GB -- we still think it is low, of course every environment is different) At 20MB log space and if your Storage manager somehow stops working, your Informix instance will also stop. I recommend increase your log space to 500MB and resize your log from 2000 to 20000 -- small logsize will cause lots of overheads. ----- Original Message ---- From: GARY PRICE <gprice@nationwideprovident.com> To: ids@iiug.org Sent: Thursday, October 18, 2007 11:51:24 AM Subject: LOGSIZE recommendation [10196] Hi. I'm currently preparing to migrate our Informix databases to new servers in the datacenter of our parent company. Also configuring an HDR pair for high availability. We've run some performance tests, including an application mass update. When this test, and the nightly update cycle run, the logs offload about once every 1-2 minutes. This seems a bit excessive, however, I realize there is a tradeoff in checkpoints/recovery time/data loss/sync'ing HPL, etc. So I thought I'd ask for some recommendations from this group. Current specifications are LOGSIZE=2000 LOGFILES=10 LOGBUFF=32 Logs are offloaded via the alarmprogram.sh using ONBAR and Netbackup to a virtual tape library, pretty quick. Note that this database/application are from a 3rd party vendor who supplied most of the database configuration parameters. Thanks for your assistance ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum. __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com