DDC Powervault Maintenance
Today we will be destroying the Apt and Utah Cloudlab block-stores to rebuild them with more redundancy.
The Powervault contains 30 4 TB drives that were previously arranged in the following manner:
- The 30 disks are split into two 15-drive RAID5 arrays, each on a separate RAID controller.
- Each array is split into two equally-sized logical drives.
- One logical drive from each array is exposed to
dbox1
, and other on each array is exposed todbox2
.
The new configuration will be as follows:
- The 30 disks are split into two 8-drive RAID6 arrays and two 7-drive RAID6 arrays. Each controller will have an 8-drive and 7-drive array.
- Each array contains only one logical drive using the entire space.
- The logical drives on the 8-drive arrays will be exposed to
dbox1
, the logical drives on the 7-drive arrays will be exposed todbox2
.
We have set aside a locked down Apt node with a RAID1 pair of 10TB drives containing all of the backed up datasets. @hibler ported all of them over earlier and has been applying syncs since. The final update has already been made as of this morning. iSCSI services have been shut down, and @hibler will be destroying the zpool on dbox1
(dbox2
is already done) before shutting down dbox1
and dbox2
. I'll re-initialize the arrays as described above, and once we confirm things look good from the dbox sides, we'll repopulate the Powervault with the datasets.