Index in a mess.
I wonder if anyone can offer me some advice so that I can clean up my Indexes for my Journal Archive ?
We currently Journal all email in and out of our organisation and store it for 1 year. All indexes are stored on the same drive but the Journal Partitions have been on several different disks each 250gb. We have been closing off the Partitions each time we reach around 240gb and opening a new one.
We have a problem now that when we try and search the Journal Archive instead of getting the results straight away we get an Archived drop down box with a list of 16 to and from dates in. This wouldn’t be so bad but these dates seem to be fairly random. For instance I have an option for 30 Nov 2007 to the 6th of Dec 2007. As we are supposed to be deleting anything over a year old I'm not sure why that is still there ? The other problem is that there are large chunks of dates just not in that list for instance I'm completely missing 1st to the 23rd of March 2009.
I have looked at the properties for the Journal Archive on the index tab and there are 16 listed and all the numbers in the "Range" column follow on without any gaps. They are also all listed as Normal with no failed items. The worrying thing is that when I look at the From and To Date columns, it list From Dates in 1979,1980 and To Dates ranging back to 2007.
What I'd like to do is completely clean up my indexes and make sure that when we search we are able to search exactly 1 year back and there are no gaps in the dates we can search for.
Is this possible ?
Any help much appreciated.
Paul
Comments
I hate saying this but the
I hate saying this but the best way to ensure your indexes are clean is to rebuild them. I know it will be a painful task but it will redo the indexes to ensure the data is consistant
What version of EV are you presently running
Liam Finn
http://www.plymouthrocklodge47.com
If you are concerned abou
If you are concerned abou Indexes and inconsistancy, you can validate the Index against the DB entries using Indexcheck
Check this TN:
http://support.veritas.com/docs/280895
Thanks for the replies.
Thanks for the replies.
Firstly we are running EV 2007 V7.5 , am I right in thinking to rebuild its just a case of right clicking and choosing rebuild index volume ? I presume I can do this in a live environment ?
I have run the Indexcheck command
indexcheck -c exist -f e:\index1 -ignorewarnings > e:\indexcheck.log
and got this at the end of it.
Finished. Checked 2477 index(es), 0 with errors, 0 with warnings
End time: 03/06/2009 14:56:29
Duration: 15 minute(s) 5 second(s)
To rebuild all for a
To rebuild all for a particular archive; go to the properties of the archive, Advanced tab and click Rebuild.
The command for Indexcheck you used is to verify locations. I would add the -c stats switch to compare the Index to the DB
the example in the doc: IndexCheck.exe -c stats -f C:\Program Files\Enterprise Vault\Indexing\1773A46CFC34... -db SQLserver2 -diff 20
I'm just wondering how I'm
I'm just wondering how I'm going to do this without taking the server down as it warns me that running live can cause corruption ? Any idea how long it will take to run ? I'm thinking maybe running overnight ?
Paul Keep in mind that an
Paul
Keep in mind that an Index rebuild could take weeks to finish...!
Cheers
Michel
www.quadrotech-it.com - All your EV Tools | www.techfreak.ch
Hi Michel, I have just ran
Hi Michel,
I have just ran Rebuilds on the first 6 indexes and they completed immediatley, although these all have 0 items listed under Total items. I presume as these were the first indexes then these are the ones that are over a year old and shouldt be available anymore anyway ? Now I've run the rebuiold command the range is now
1-0
6345131-0
7893258-0
9391848-0
10915432-0
12405512-0
Oh, Paul. You could use
Oh, Paul.
You could use FederatedSearch:
http://seer.entsupport.symantec.com/docs/292820.htm
I know it states that it was added in 8.0 SP1, but I believe it works for previous version, too.
This should search more than one index volume, so that you will get a unified result set.
Cheers
www.quadrotech-it.com - All your EV Tools | www.techfreak.ch
Hi, I don't see a problem
Hi,
I don't see a problem with the indexes from a quick read of the above. The data has been stored across several index volumes and so is showing you the each of these volume. The dates I think are the dates on the emails rather than the archived date so that might explain the data overlap. So if you look you will see emails with those dates. EV hasn't got it wrong. I don't think rebuilding will change the data.
Since it's a journal archive then it is already using federated search, that is why you are seeing results from more than one index volume.
So I think some have rebuilt quickly because infact they are empty as all the items from them have been expired.
Like I say kind of an educated guess based upon a quick read.
Mike Bilsborough
Director,Enterprise Vault Engineering Support
more information
So by default when you indexes are 5 or less indexvolumes then it will just do the search across those 5 for you. If you have greater than 5 then it basically is trying to say hey this is just too much data to search, can you be a bit more specific and so it lists the particular index volumes for you to choose and then when you choose one it will then do the search for that particular index volume.
Mike Bilsborough
Director,Enterprise Vault Engineering Support
Paul What is your use case
Paul
What is your use case for searching the Journal using "standard" EV Search?
Maybe, you could benefit from Discovery Accelerator?
Cheers
www.quadrotech-it.com - All your EV Tools | www.techfreak.ch
ok even more info
Hi,
So as Michael says you can set MaxVolSetsToSearch to a higher number but probably that's not ideal.
Also I think you have chosen to rebuild based upon the indexvolume tab. The thing with that is that you'll still end up with the same number of indexvolumes and so still too many to be easily searched. If instead you choose rebuild on the 'advanced tab' that will effectively delete all your existing index volumes and start again from scratch so then it may just end up with 3 or so indexvolumes is enough to hold all your active data for example.
So assuming your old index volumes are empty I think there is a slight bug in that EV is listing index volumes that are actually empty.
Mike Bilsborough
Director,Enterprise Vault Engineering Support
Thanks again for all the
Thanks again for all the replies.
The reason for the searching is mainly for disciplinary and FOI requests from outside of our organisation and are performed by a Data Protection Officer.
The DPO just gets a little confused when he's seeing dates back to 2007 within the search when he knows we only store email for 1 year, and also some dates he has to search on simply dont appear in the list.
Your right I was looking at the Index volume tab. I've now kicked of a complete rebuild but done it from the Advanced tab, we'll see how many weeks it takes !
Thanks again for all the help.
Weeks was right ! Still if it
Weeks was right ! Still if it cleans everything up then its worth it.
Percentage completed: 0%
Estimated time of completion: 17/06/2009 16:24
Wow 2 weeks you are so
Wow 2 weeks you are so lucky
I have to rebuild mine in the comming months but it will take me 6 months at least to get mine rebuilt
BTW. If one if the posts is your solution please mark it as such to keep only open issues unmarked
Liam Finn
http://www.plymouthrocklodge47.com
My reindex has finished
My reindex has finished ! Quicker than expected. All looks good now with the dates in that I have a complete range from up to a year ago. I just have one small niggle.
When I search my journal under the Archived dropdown list I have all the correct dates but they are not in date order, ie my first date range is
13 Oct 08 til 30th Oct 08, but then the next in the list is 02 December 08 - 17 December 08 . Is there anyway to get these to apperar in the correct order so I dont have to browse through the entire range to get the dates I'm looking for ? Or even better get rid of the drop down and just have it automatically span the entire index ?
Thanks
Paul
Nope afraid not I recall this
Nope afraid not I recall this being mentioned various times but obviously this has not been rectified or I do not recall seeing anything statung that this has been done.
To get it to not give you the drop down you just need to set FederatedSearchMaxVolSets in webapp.ini to a larger value than the number of index volume sets you have. (Default is 5)
EV Backline Technical Support Engineer APJ Region
Paul Could you mark this
Paul
Could you mark this thread as solved if you're satisfied with this answer?
Cheers
www.quadrotech-it.com - All your EV Tools | www.techfreak.ch
Would you like to reply?
Login or Register to post your comment.