Hi,Since you probably don't have enough space for backup for such a large 
table, test it on a small one and rowcount before and after.We work extensively 
with mob and there are several issues with it that we had to work around.1.Mob 
and ttl - region server crashes after ttl archiver runs.2.Mob and snapshot - 
didn't work in cdh 5.7.13.Mob compaction- currently performed only by the 
master.if your flushes aren't big enough , compaction will never end.4.Deletes 
- major mob compaction is the only one doing mob delete,it is also done by the 
master, on the entire data( it is not scheduled )If you have deletes, the will 
never actually be deleted.
Regards,Meirav
Sent from my Samsung Galaxy smartphone.
-------- Original message --------From: Jean-Marc Spaggiari 
<[email protected]> Date: 9/21/18  01:09  (GMT+02:00) To: user 
<[email protected]> Subject: Re: Can MOB be enabled for existing table? 
Hi Andrea,

I will say that the easiest way to know is to try it... I'm guessing that
it should work. When HBase will compact the file, it will figure that some
fields are bigger than the configures MOB threshold and will move them
under the MOB umbrella. But this has to be tested. Let us know.

Thanks,

JMS

Le jeu. 20 sept. 2018 à 08:26, Andrea Fonti <[email protected]> a
écrit :

> Hi,
>
> I’m currently running HBASE on a Cloudera CDH 5.8.4 distribution
> (hbase-1.2.0+cdh5.8.5+263), is it possible to enable MOB for a column of an
> existing multi terabyte table?
>
> If the answer is affirmative what will hbase do with the existing data?
> Will it rewrite with MOB files the next major compaction?
>
> Thank you
>
> Andrea

Reply via email to