Skip to content

Burned out on 24/7 on-call, considering hosting industry exit

General Discussion by SingaporeRep 8 replies 513 views
8 #1
Fio --name=burnout --readwrite=randread --bs=4k --iodepth=64
burnout: (g=0): rw=randread, bs=(R) 4096B-4096B
iops        : 2473
lat mean    : 25894.31 msec  # my reaction time at 3am
cpu         : 98.5% usr,  0% sleep

Decent disk IO, meh network. My latency to happiness is worse. 8 years NOC to SRE, now "platform engineer" same pager. Considering hetzner.com or just leaving.

  • Stay: $good, pager, 60hr weeks
  • Move: $less, maybe pager, unknown
  • Exit: $???, no pager, unskilled
fio, iperf, geekbench. results or gtfo.
#2

Virtualization tax isn't just CPU overhead; it's the cognitive load of context-switching between "normal life" and "production is down." OpenVZ was worse because you couldn't debug the host properly; at least with KVM you have full isolation between your sanity and the guest's chaos. Eight years is enough to know the patterns; the 3am pages are never novel, just expensive. The industry won't fix its on-call culture; cgroups can't limit management expectations.

virsh list --all | wc -l: 47
#3

I felt this in my bones and I need to say it all at once because I have been there and I am still there sometimes and the thing is you think it gets better when you move up or sideways and sometimes it does for a little while and then the same patterns emerge and you are the only one who knows that one legacy system and the documentation is in someone's head and that someone left and now it's your undocumented baby and you wake up at 2am not even from the pager but from the anticipatory anxiety that the pager might go off and your partner is used to it now and that is somehow worse and I tried the boundary thing and my manager said "we're all in this together" and I said "then why am I the only one getting paged" and he had no answer and I stayed another year and then I left for a smaller host and it was the same and then I left again and now I am at RackNerd and it is slightly less bad

#4
SingaporeRep said:
My latency to happiness is worse

+--------+ +--------+ +--------+
| AS64512 |---| AS64513 |---| AS64514 |
+--------+ +--------+ +--------+
| | |
[you] [pager] [exit?]

RFC 1122: "Be liberal in what you accept" - bad advice for career boundaries.

I had ASN before I had degree. Anyone without AS cannot understand peering relationships, cannot understand that your upstream (management) is single-homed to burnout. I left after prefix hijack of my personal time. Now consult on BGP only, no on-call. My advice: get autonomous, literally.

+--------+
| AS you |---[no pager]
+--------+

#5

I feel you. Have you tried tracking which pages actually needed immediate response? In my experience, 70% of 3am pages could wait until morning; the client just wanted acknowledgement. The "dumb" questions at 2am are often people in other time zones who don't know it's night for you. I gently corrected a manager once by showing him the data: we were paying wake-up premium for non-urgent tickets. He actually changed the escalation policy. Try the data approach; it sometimes works.

#6

Guys this thread hits hard. I was NOC before I pivoted to reseller, now I just get angry client tickets instead of pages. Different stress, same sleep loss. Any leads on roles with actually healthy on-call? Or margins good enough to hire coverage? My friend left for fintech, says is worse - more money, more pager.

Thinking about exit too but I recieved good advice once: don't quit from burned, quit to something. Still searching what that is. 🙏

your margin is my opportunity
#7

What's the actual pay gap between stay and move? I need numbers.

Vive la résistance... électrique
#8

The anticipatory anxiety hit me too. I stopped sleeping before on-call weeks.

phở at 3AM, deploy at 4
#9

Hetzner won't fix the pager. Different logo, same 3am kernel panic.

Post a reply

You need an account to reply. Log in or register to join the conversation.

Post reply Preview Save draft