Difference between revisions of "Backups"

From wiki
Jump to: navigation, search
(Created page with "Jason Cormie of IT Services supplied the following details The volume is raided across 348 disks (mix of SSD,SATA, SAS with 29 being hot spares), and asynchronously mirrored...")
 
Line 1: Line 1:
 +
= Introduction =
 +
 +
It is very important to note that backups are usually only 7 days old.
 +
 +
In the case of accidental deletions, you must inform immediately as it could take several days to load up the snapshot, consider that it is many TB in size.
 +
 +
= What it entails =
 +
 +
* A file system has to be created from the snapshot.
 +
* this is loaded onto the SAN
 +
* The SCSI bus needs to be rescanned
 +
* this snapshot-rendered-to-filesystem must then be mounted.
 +
 +
= Background information =
 +
 
Jason Cormie of IT Services supplied the following details
 
Jason Cormie of IT Services supplied the following details
  

Revision as of 13:04, 3 April 2017

Introduction

It is very important to note that backups are usually only 7 days old.

In the case of accidental deletions, you must inform immediately as it could take several days to load up the snapshot, consider that it is many TB in size.

What it entails

  • A file system has to be created from the snapshot.
  • this is loaded onto the SAN
  • The SCSI bus needs to be rescanned
  • this snapshot-rendered-to-filesystem must then be mounted.

Background information

Jason Cormie of IT Services supplied the following details

The volume is raided across 348 disks (mix of SSD,SATA, SAS with 29 being hot spares), and asynchronously mirrored to our secondary datacentre in Maths.

Data scrubbing runs in the background regularly to detect any bad blocks, before they become an issue.

Snapshots for rollback are also mirrored to the secondary DC.

We can do synchronous mirroring, but it would have a performance impact as you’d have to wait for commits from the far side every time you wrote.

Note: synchronous mirroring is not enabled.