[Glass] Tranlogs space on Linux

Dale Henrichs via Glass glass at lists.gemtalksystems.com
Thu Dec 17 17:58:55 PST 2015


On 12/17/2015 11:44 AM, BrunoBB via Glass wrote:
> Dale,
> Why i need so many tranlogX.dbf to restore a backup ?
When you make a backup, you need to keep all of the tranlogs that were 
created after the backup was made ... otherwise you won't be able to 
restore from your backup and the replay tranlogs to restore your system 
to the state prior to crash ...

If you make a new backup you can throw all of the tranlogs associated 
with the old backup and start keep tranlogs for the new backup ...

> I'm reading chapter 8 "To Initiate a Checkpoint"
> If i do:
> System startCheckpointSync.
> The tranlog will start from 0 again ?  or it is just a like a commit without
> realising locks ?
no the tranlog file names increase with each new tranlog ... the stone 
keeps track of the tranlog numbers it has written and the ones that it 
needs to restore ...

When you restore from a backup, the last tranlog file written is 
recorded, so the stone will know what tranlog is needed to restore post 
> Page 54:
> "At any point, the method Repository>>oldestLogFileIdForRecovery identifies
> the oldest log file needed for recovery from the most recent
> checkpoint,...."
> Is System startCheckpointSync this checkpoint ?
Yes that method starts a checkpoint ...

Keep in mind that the tranlogs are used for recovery in two cases:

   1. the stone crashes and you restart the tranlog - which starts at the
       most recent checkpoint and then recovers from tranlogs to get the 
       recent commit.
   2. the stone crashes and the extents are corrupted, so you have to 
       from backup in which case after restoring from the backup, you 
will need
       to restore from tranlogs kept since the backup was made 
(including the tranlogs
       needed to restore to the checkpoint of the backup)


More information about the Glass mailing list