Hi All,
I'm just making sure this is a solid enough plan before I execute.
I have a 3 node, 5 disk(each) cluster with ceph storage. currently the cluster consists of (combined) 3 2TB NVME and 15 2-3 TB spinning disks. I just got approval last week to convert all 15 spinning disks to 2TB ssds (and add 3 more SSDs) but I can't do it until the budget resets at the end of the year. My luck; i lost a drive yesterday, exactly one week after my conditional approval. We're certain it's dead, the OSD keeps crashing and the node is spitting out some nasty looking IO errors for the affected disks.
To my crazy question; how bad of an idea is it to replace the failed spinning disk with an SSD today? I'm aware that I wont see the performance benefits of the SSD in this configuration. My thinking is why buy a new spinning disk today just to completely replace it for an SSD in 3 months; I'll have no use for these spinning disks after the upgrade. so instead of buying 18 SSD in January we'll only buy 17 (or 16 if we loose another one before then). I'm unsure if there are any performance or stability issues when mixing drive types like this. Maybe there is a completely different type of issue when mixing disks that i'm ignorant to.
It will only be until the end of the year but if this a risky operation we'll replace the failed disk with a comparable and upgrade all the SSDs as one project in January.
Thoughts? (feel free to just tell me i'm nuts...
)
I'm just making sure this is a solid enough plan before I execute.
I have a 3 node, 5 disk(each) cluster with ceph storage. currently the cluster consists of (combined) 3 2TB NVME and 15 2-3 TB spinning disks. I just got approval last week to convert all 15 spinning disks to 2TB ssds (and add 3 more SSDs) but I can't do it until the budget resets at the end of the year. My luck; i lost a drive yesterday, exactly one week after my conditional approval. We're certain it's dead, the OSD keeps crashing and the node is spitting out some nasty looking IO errors for the affected disks.
To my crazy question; how bad of an idea is it to replace the failed spinning disk with an SSD today? I'm aware that I wont see the performance benefits of the SSD in this configuration. My thinking is why buy a new spinning disk today just to completely replace it for an SSD in 3 months; I'll have no use for these spinning disks after the upgrade. so instead of buying 18 SSD in January we'll only buy 17 (or 16 if we loose another one before then). I'm unsure if there are any performance or stability issues when mixing drive types like this. Maybe there is a completely different type of issue when mixing disks that i'm ignorant to.
It will only be until the end of the year but if this a risky operation we'll replace the failed disk with a comparable and upgrade all the SSDs as one project in January.
Thoughts? (feel free to just tell me i'm nuts...
)