The index hang at the end of a group could be down to the parallelism you have set on your NetWorker server client definition. If by default it is 4 (old default, new default is 12) from NetWorker 7.4 onwards the index backups count as a session for the NetWorker server, so if you had a group with 20 clients in it would only backup 4 indexes concurrently and the other 16 would feed in as index backups complete. This is of course providing you are not also backing up the NetWorker server. A reasonable parallelism is 12-16 - this should allow you index backups to go through without a long delay being seen.
A lot of people create an Index group, link one of each unique clients to this group and use savegrp -O Index to do an index backup only and then a bootstrap to tape. This has an advantage of in a DR situation you will have all the information on one or two tapes (instead of many) and helps recovery of the NetWorker server information to be quicker (as you do not have to load or have many tapes available). Downside is you must have each client linked to the group otherwise that clients index will not be secured, so when creating new clients you must remember to link to the Index group.
Have a look at the Parallelism for the NetWorker server client definition and increase to see if it makes any difference.
Bill covered most of the pain points, so just one more: keep in mind that indexes have retention also, so when you backup the indexes to the same tape of the backup that just run, the indexes will expire along with the saveset. If you send the indexes to a separated pool, you may have to keep old indexes for a longer time, due to other ones that need to be kept. Not a big issue as indexes are not that larger most of time, but I've seem cases where it was really tape consuming.
Just one clarification with Bill's explanation regarding having a separate group for the index and bootstrap: is the schedule for backing up still the same with the group schedule? Meaning after a group session completes, we still need to immediately backup the index but only to different group or it's done on a different schedule, say once daily?
Sorry, Bill and Daniel, but I just got a bit confused based on Daniel's input: "If you send the indexes to a separated pool, you may have to keep old indexes for a longer time, due to other ones that need to be kept." Let me just clarify again whether my understanding of the index saveset is correct.
My understanding of Index backups is that it contains ALL the group's saveset content since the particular group was created (assuming we are not doing any purging). Is this correct? If so, if I am going to do a daily index save using savegrp -O, why should I still have to have a longer retention for the index saveset when tomorrow when I do another index save, it will just save all old data again together with the additional new ones. Therefore, if my understanding is correct, the older index savesets can actually have a retention period of just 1 day because it should still contain the saveset data even if my retention for the data itself is 10 years. Again, please correct me here if my understanding is wrong.
Hi, using savegrp -O is usually done on a daily basis, not after each group. You can either let each group secure the indexes also or select no index save option which will not try to secure the index at the end of the group and then you will only backup the index when savegrp -O is run. Either is acceptable depending upon your reasons for wanting to run savegrp -O (some people with large indexes for clients do this because of the time and apace required, others just for DR purposes so they can recover quickly from one or two tapes.
Well, I think I made a little confusion in my answer
When backing up indexes as part of the group backup it uses level 9 backup, and you're right, savegrp -O will save the full indexes (never used in production environment tough), thus allowing you to have minimum retention. I kind of mixed the levels in my mind before answering ... I think I'm getting old..
masonb
445 Posts
0
February 5th, 2010 01:00
The index hang at the end of a group could be down to the parallelism you have set on your NetWorker server client definition. If by default it is 4 (old default, new default is 12) from NetWorker 7.4 onwards the index backups count as a session for the NetWorker server, so if you had a group with 20 clients in it would only backup 4 indexes concurrently and the other 16 would feed in as index backups complete. This is of course providing you are not also backing up the NetWorker server. A reasonable parallelism is 12-16 - this should allow you index backups to go through without a long delay being seen.
A lot of people create an Index group, link one of each unique clients to this group and use savegrp -O Index to do an index backup only and then a bootstrap to tape. This has an advantage of in a DR situation you will have all the information on one or two tapes (instead of many) and helps recovery of the NetWorker server information to be quicker (as you do not have to load or have many tapes available). Downside is you must have each client linked to the group otherwise that clients index will not be secured, so when creating new clients you must remember to link to the Index group.
Have a look at the Parallelism for the NetWorker server client definition and increase to see if it makes any difference.
Regards,
Bill Mason
dpinink_silva
2 Intern
•
724 Posts
1
February 5th, 2010 06:00
roose
6 Posts
0
February 6th, 2010 20:00
Thanks Bill and Daniel!
Just one clarification with Bill's explanation regarding having a separate group for the index and bootstrap: is the schedule for backing up still the same with the group schedule? Meaning after a group session completes, we still need to immediately backup the index but only to different group or it's done on a different schedule, say once daily?
roose
6 Posts
0
February 8th, 2010 00:00
Sorry, Bill and Daniel, but I just got a bit confused based on Daniel's input: "If you send the indexes to a separated pool, you may have to keep old indexes for a longer time, due to other ones that need to be kept." Let me just clarify again whether my understanding of the index saveset is correct.
My understanding of Index backups is that it contains ALL the group's saveset content since the particular group was created (assuming we are not doing any purging). Is this correct? If so, if I am going to do a daily index save using savegrp -O, why should I still have to have a longer retention for the index saveset when tomorrow when I do another index save, it will just save all old data again together with the additional new ones. Therefore, if my understanding is correct, the older index savesets can actually have a retention period of just 1 day because it should still contain the saveset data even if my retention for the data itself is 10 years. Again, please correct me here if my understanding is wrong.
masonb
445 Posts
1
February 8th, 2010 00:00
Hi, using savegrp -O is usually done on a daily basis, not after each group. You can either let each group secure the indexes also or select no index save option which will not try to secure the index at the end of the group and then you will only backup the index when savegrp -O is run. Either is acceptable depending upon your reasons for wanting to run savegrp -O (some people with large indexes for clients do this because of the time and apace required, others just for DR purposes so they can recover quickly from one or two tapes.
Regards,
Bill Mason
dpinink_silva
2 Intern
•
724 Posts
0
February 8th, 2010 05:00
Well, I think I made a little confusion in my answer
When backing up indexes as part of the group backup it uses level 9 backup, and you're right, savegrp -O will save the full indexes (never used in production environment tough), thus allowing you to have minimum retention. I kind of mixed the levels in my mind before answering
... I think I'm getting old..