Hi
We are investigating what High-Availability and Disaster-Recovery best practices we can implement for our on-premises solution for Jira Software.
Any guidance will be greatly appreciated
FD
1. if you can have datacenter license you can have better HA solution as recommended by Atlassian.
2. If you have SAN storage you can try this. We use it and it works fine for us.
- Store JIRA installation and home directories in SAN storage which is mounted to JIRA app server
- sync same mounts (Storage LUNs) to DR premises
- you need to have another server in DR site
-once primary site failure, mount DR mount points to DR server and continue.
- This has a downtime, not a smooth failover
- Do same for DB server as well
>Store JIRA installation and home directories in SAN storage which is mounted to JIRA app server
No! Do not do that. It will appear to work for relatively small low use Jira systems but will fail at scale. It's arguable for the installation directory, but there are parts of the home directory that you absolutely do not place on SAN storage (and others that you should)
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
We configured our Atlassian systems in SAN storage, but SAN storage has same disk performance as same as local disks. This is not a NFS mount. It is a SAN mount which is connected via fiber cable from the server to SAN storage. We have almost 1800 active jira users in out JIRA instance with 700k JIRA issues. According to disk write speed tests, we do not see any performance issue. The only thing is you should have high-performance infrastructure.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Then your performance tests are wrong, or you are using really under-performant local disks.
The performance of your disks is not that important though, you're measuring the wrong thing. You need to measure your Jira performance and the error rates you are getting (these should be zero on local disks, but won't be on a SAN)
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Nic,
I also thought the same before we implement this. We had normal disks for the local disk but we use SSD for SAN storage. I think it might be the reason for performance. However, if you have the latest HW for SAN, it can provide similar performance as same as local disks. We shared our performance test data with Atlassian as well before we implement this solution. Apart from the performance, there is a risk when storing data in a local disks.
However, we should not use NFS mount points to store data.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.