So, I came up with and idea to automate the process of doing snapshots at regular intervals. There is nothng better than being able to take a task off your weekly calandar. So I went into a CVM and created an automated task to do the snapshots.
I ran into a problem though. Should that CVM ever go down the auto-snapshot function that I created would stop working.
So I ventured to do the next best thing. Create a VM, which has the ability to move around, that would ssh into an available CVM and run the task for me. But the problem with that one is I am taking up resources that I would not need to use otherwise.
So why not do the next best thing right? Take the script and copy it to all the CVMs and schedule the task? I'm sure you can see the problem immediately. Now I'm creating 3 snapshots instead of the one. I could fix this by doing some error checking though. Have the script look at the snapshots and if one exists between a certain date and time range skip it. Turns out this doesnt work quite like I had hoped it would for a few reasons.
So, it needs to be integrated into or inside of Ergon. Basically make it the type of object as a health check, one that is replicated and ran across all of the nodes and managed by Acropolis. Maybe even a function of Cerebro.
Solved! Go to Solution.
Are you trying to snapshot the CVM itself? There shouldn't be a reason to do that because all the data it contains is replicated to other CVM's in the cluster. If it has to be recreated, it will gather data back after it's re-added to the cluster.
Yeah, lol, i figured out very quickly that the CVMs were fairly protected from catastrophic devistation.
Yes. It would, if I had another cluster to do DR to. Problem with that is money. What spawned this is that we had a VM that filled up, so I expanded the disk which caused windows not to see the disk anymore. Needless to say that VM was toast. It was a very low level machine that was a convert from VMWare, which we have had stability issues with VMWare converts in the past. So it would have been nice if I had a valid snapshot. Its not that I cared about that machine. I was just thinking if it happened to a production VM, I would have to rebuild it from the ground up. And no, cloning the machine or using its disk image did not work. I spend a few hours trying to get it back up and running somehow.
Really? I thought you had to have a replication target to set up the PD. I think I remember when I initially tried to set that up back in Nov/Dec time frame I didnt have a replication target so I couldnt set it up. I will have to play with that now.