oO.o's Neverending Tech Blog

I’d love to have some old Apple server to play around, but they are quite rare up here.

And some server hardware would give me good excuse to get server rack for myself. :smile:

1 Like

It’s been a great investment for me. I think I built it from used parts 5 or 6 years ago. Interestingly, this particular model is the only Mac to ever have an OOBM capability.

I can’t virtualize anything newer than macOS Mojave on it, so it will be replaced with Mac Minis soon. Just have to find the time.

1 Like

Just out of curiosity, how many times in a year you have to drive to the data center?

1 Like

I’ve been twice since 2016 including the initial install.

I’ve used the on site support 3 times to perform hard resets when things have locked up.

1 Like

Oh that’s pretty good :smile:

1 Like

Wow, so since my ESXi host is on the verge of death, I figured it was a good time to set up iscsi and start migrating VMs off of the local storage.

Got it set up fine, but when I went to format it, it would fail with some ambiguous errors. The fix ended up being to enable flow control on the switch. Been a long time since I’ve even thought about flow control. I guess it’s still a thing…

1 Like

Yeah, I’ve been happy with it. The on-site support is $100/hour in 1/2 hour increments, so you really don’t want to be calling them up every week to reboot something for you. In my experience, there are 2 things that help immensely with this.

First is IPMI or any OOBM you have that has a KVM functionality. Usefulness is kind of obvious. You can edit bios remotely and get out of crashed operating systems.

Second is scheduled reboots on network hardware. Whenever I’m doing extensive network config, I’ll schedule a reboot ahead of time. The worst case is that I’ll lock myself out until the scheduled reboot occurs and then the previous config will load and I’m back to where I was. This is why you commit and then save config on routers and switches. If the commit locks you out, the last saved config will load on reboot. I absolutely have locked myself out of my switches because of bugs or user error and the scheduled reboot saved me $50 each time.

Also worth mentioning, I have a netboot.xyz usb in each machine. While this is not strictly necessary because I can mount ISOs remotely via IPMI, it is much much faster.

4 Likes

Yeah IPMI is good. I wouldn’t mind having it even in my file server that is next to my desk. Not saying I’m lazy to start connecting keyboards and displays to it if I f up something over SSH but, well I’m lazy… :smile:

What hw setup you are planing to use with Mac minis? 8 cores and 64gb + 10gig nic or something more casual?

1 Like

Good question. Right now I own a 2012 quad-core Mac Mini and a current-gen Mac Mini with 10GbE, both of which are officially supported ESXi hardware with the caveat that ESXi cannot use the SSD on current-gen Mac Minis because of Apple’s encryption nonsense.

I had planned to use the new Mac Mini has my Apple workstation at home, but since covid, that space is being filled by the Mac Pro I have from work. But I think I’ll still want to use it as a workstation once the Mac Pro goes back to the office. So I think I’ll buy a few older Mac Minis and make a little cluster out of them. The only thing I’ll really miss is 10GbE, but I’ll deal with it.

1 Like

:thinking:

2 Likes

whats that run to hit a power button?

nvm

2 Likes

Sometimes they’re nice and don’t charge me though.

2 Likes

I really like juniper because it will apply the change then revert if you dont confirm in like 30 seconds, prevents the reboot if you fuck up and lock yourself out (Really surprised no cisco gear i have ever used had anything like that)

1 Like

Got this snapshot error across 3 completely different FreeNAS servers last night. Weird.

Snapshot Task For Dataset "pool1/srv" failed: cannot create snapshot 'pool1/srv@auto-2020-11-01_01-00': dataset already exists

2 Likes

DST change :stuck_out_tongue:

2 Likes

Oh lol. Totally forgot that happened last night.

Oh, shit. That explains a lot.

I was wondering why all our monitoring threw errors last night.

3 Likes

I guess that’s why we should use UTC. Weird thing is I don’t remember this happening in previous years.

2 Likes

Y2k20 too strong

4 Likes

Unexpected shell behavior:

set -o pipefail
{ command1 || command2; } | command3

If command1 succeeds but command3 fails, command2 is attempted.