A client's server (a Dell r630) began experiencing issues (originally due to a faulty switch, but it led to new hardware purchases). It was running Proxmox, which I installed about three years ago, featuring SSD drives and 128 GB of RAM. A new server was set up around four months ago, but a water leak from the floor above flooded the rack, rendering the new hardware unusable. In a rush yesterday morning, I quickly reconfigured the "old" r630, installing FreeBSD 14.0-RELEASE and transferring VMs from the latest backup using zfs-send/receive. After launching the VMs on bhyve with some minor adjustments, everything was back in production. Just got a call saying the performance has "drastically improved." Previously, a specific end-of-month process on the r630 with Proxmox took about 30 minutes, while the new (flooded) server cut that down to 15 minutes, much to their delight. This morning, the same process only took 8 minutes, to their great surprise. I'm curious to investigate, but can't quite pinpoint how this is possible: my suspicion is that the process is mainly I/O bound, and the nvme driver in bhyve is exceptionally efficient in this aspect. For once, I'm taking credit for something I never expected. :-)
#TechLife #DataRecovery #ServerManagement #ITChallenges #bhyve #Proxmox
Stefano Marinelli /
npub14c…j0nj8
2024-01-26 14:07:58
Author Public Key
npub14calwd6xg349ahf3nnhhyqem2w2e3gs66p7zctz2sna74u3tsddq7j0nj8Published at
2024-01-26 14:07:58Event JSON
{
"id": "7047841576de2dadd06ab6ae52c7bc2db6c7fd9588298ba65f2c5ff55cb869a1",
"pubkey": "ae3bf73746446a5edd319cef72033b539598a21ad07c2c2c4a84fbeaf22b835a",
"created_at": 1706278078,
"kind": 1,
"tags": [
[
"t",
"techlife"
],
[
"t",
"datarecovery"
],
[
"t",
"servermanagement"
],
[
"t",
"itchallenges"
],
[
"t",
"bhyve"
],
[
"t",
"proxmox"
],
[
"proxy",
"https://mastodon.bsd.cafe/users/stefano/statuses/111822640155134018",
"activitypub"
]
],
"content": "A client's server (a Dell r630) began experiencing issues (originally due to a faulty switch, but it led to new hardware purchases). It was running Proxmox, which I installed about three years ago, featuring SSD drives and 128 GB of RAM. A new server was set up around four months ago, but a water leak from the floor above flooded the rack, rendering the new hardware unusable. In a rush yesterday morning, I quickly reconfigured the \"old\" r630, installing FreeBSD 14.0-RELEASE and transferring VMs from the latest backup using zfs-send/receive. After launching the VMs on bhyve with some minor adjustments, everything was back in production. Just got a call saying the performance has \"drastically improved.\" Previously, a specific end-of-month process on the r630 with Proxmox took about 30 minutes, while the new (flooded) server cut that down to 15 minutes, much to their delight. This morning, the same process only took 8 minutes, to their great surprise. I'm curious to investigate, but can't quite pinpoint how this is possible: my suspicion is that the process is mainly I/O bound, and the nvme driver in bhyve is exceptionally efficient in this aspect. For once, I'm taking credit for something I never expected. :-)\n\n#TechLife #DataRecovery #ServerManagement #ITChallenges #bhyve #Proxmox",
"sig": "4ed8cce9682b454f3ba08974e7b76cf6fae6b339bcbdb1ff25bfd46cd00a89623059e6edf541c612e65b5a6e07a47dd26c8c22d8d0e049437e1884b08a96b948"
}