HDR and countinuous l.l. backup
Posted in 2003
Topics: High Availability & Replication, Logging & Checkpoints
Hello, we are going to start Informix HDR in asynchronous mode. Now we are doing continuous l.logs backups on the (primary) server. Couldn't we end up with the situation when logical log on the primary server is written to tape and freed before replication to the secondary? Or it's prevented by doing synchronous checkpoints on both servers in the pair? Unbuffered logging is being used for all databases. Thank you for advice. Best regards, Laurynas Barauskas IS administrator EPS, ISD AB Vilniaus bankas ---- Gedimino pr. 10, Vilnius, LT-2600 LITHUANIA Tel.: +370 5 2682624 Mob.: +370 686 30502
Hi Laurynas, HDR transfers the LOGICAL LOG buffers (not the logical logs) when they are written to the secondary ...Its only during checkpoints when they both have to be synchronized (If you are running on ASYNC mode)...so say Primary is on Log 12 and on the Secondary its still on Log 10 then when a checkpoint happens on the primary then the secondary has to rollforward to the same log page where the primary is. So, I guess in your case you shouldn't worry about the logical logs being marked as free after the backup ...:-) Thanx much, Rajib Sarkar Advisory Support Engineer (Wells Fargo Bank) IBM Data Management Group Ph : (602)-217-2100 Fax: (602)-217-2100 As long as you derive inner help and comfort from anything, keep it -- Mahatma Gandhi "Laurynas Ba...." <lbarauskas@vb.lt To: ids@iiug.org > cc: Sent by: Subject: HDR and countinuous l.l. backup [378] forum.subscriber@ iiug.org 02/13/2003 08:26 AM Hello, we are going to start Informix HDR in asynchronous mode. Now we are doing continuous l.logs backups on the (primary) server. Couldn't we end up with the situation when logical log on the primary server is written to tape and freed before replication to the secondary? Or it's prevented by doing synchronous checkpoints on both servers in the pair? Unbuffered logging is being used for all databases. Thank you for advice. Best regards, Laurynas Barauskas IS administrator EPS, ISD AB Vilniaus bankas ---- Gedimino pr. 10, Vilnius, LT-2600 LITHUANIA Tel.: +370 5 2682624 Mob.: +370 686 30502
Hi, Couldn't you end up ... ? Theoretically : yes. Practically : no. During normal operation, it should not happen, that your Primary "runs away" and your Secondary can't follow up. After all, you adhered to our recommendations, and both machines are the same size/speed/IO rates/configuration, etc. :) If normal operation is interrupted (e.g. network problem), then your best hope is, that things get back to normal before the logs on Primary wrap around ... As I said before, having bigger and/or more logical log files will increase the period where "online-synchronisation" is possible ... Synchroneous HDR replication may slow down the Primary. Not sure if that's desired, but you might try. But if your Secondary is significantly slower than your Primary - and that over a prolonged period of time, than you've got a problem anyway. And that should be solved first, I guess. Regards, Martin -- Martin Fuerderer IBM Informix Development Munich Data Management Solutions "Laurynas Ba...." <lbarauskas@vb.lt> Sent by: forum.subscriber@iiug.org 13.02.2003 16:26 To: ids@iiug.org cc: Subject: HDR and countinuous l.l. backup [378] Hello, we are going to start Informix HDR in asynchronous mode. Now we are doing continuous l.logs backups on the (primary) server. Couldn't we end up with the situation when logical log on the primary server is written to tape and freed before replication to the secondary? Or it's prevented by doing synchronous checkpoints on both servers in the pair? Unbuffered logging is being used for all databases. Thank you for advice. Best regards, Laurynas Barauskas IS administrator EPS, ISD AB Vilniaus bankas ---- Gedimino pr. 10, Vilnius, LT-2600 LITHUANIA Tel.: +370 5 2682624 Mob.: +370 686 30502