3 ms·

Weird. Not sure why i would like to update/change a live config. That would probably not mean that there is an Infrastructure code in git.

popotamonga 6y ago
Well i have 1M+ users and still ssh into the VMs and live-change nginx configs. Once in a while the reload fails because i messed up something.
- purerandomness 6y ago
  Always run `nginx -t` before reloading to test the new config.
- battery423 6y ago
  And your solutio is to log into a VM through ssh instead of fixing it? I don't mean this as offending but my personal action is to fix the underlying issue. If you add more tests etc. you will have to do that less often until it doesn't is the norm anymore.
- mad182 6y ago
  I have over 100k uniques daily, 7 servers and I also change nginx config manually via ssh.
  - battery423 6y ago
    But why? Even if you don't use kubernetes or docker images, you do wanna have it in code right? Having something like ansible would do the same thing but allows you to have your config backuped by git and allows you to recreate the env without any issues. How do you keep your mean time to recovery low? How do you document your setup? How do you automated tests if you change your config manual?
    - mad182 6y ago
      Why not? I have done this for 10+ years and it's still working flawlessly, I don't see a reason to add more complexity to something that is working fine for my needs. I don't have to change config often, maybe once a year or so.
      - battery423 6y ago
        " I don't see a reason to add more complexity to something that is working fine for my needs" valid point though. I would do it due to: - keeping 'meantime to recovery' low due to having my configs in git -> fast recovery - using something like ansible for your small case to have it in git (instead of a shellscript) -> backup - perhaps a jenkins pipeline with testing the config before applying it automatically -> for quality - automatisation for having fun with it, seeing what is out there besides ssh and manuel editing At the end of the day, i assume that this automatisated process is also less dependend from myself. So i'm not sure who is doing things when you are on holiday or sick. I'm now having a small k8s cluster at home. I have never had anything which allows me to have my whole setup in git so easy with autohealing, domain management and high availability.

popotamonga 6y ago

Well i have 1M+ users and still ssh into the VMs and live-change nginx configs. Once in a while the reload fails because i messed up something.

purerandomness 6y ago
Always run `nginx -t` before reloading to test the new config.
battery423 6y ago
And your solutio is to log into a VM through ssh instead of fixing it? I don't mean this as offending but my personal action is to fix the underlying issue. If you add more tests etc. you will have to do that less often until it doesn't is the norm anymore.
mad182 6y ago
I have over 100k uniques daily, 7 servers and I also change nginx config manually via ssh.
- battery423 6y ago
  But why? Even if you don't use kubernetes or docker images, you do wanna have it in code right? Having something like ansible would do the same thing but allows you to have your config backuped by git and allows you to recreate the env without any issues. How do you keep your mean time to recovery low? How do you document your setup? How do you automated tests if you change your config manual?
  - mad182 6y ago
    Why not? I have done this for 10+ years and it's still working flawlessly, I don't see a reason to add more complexity to something that is working fine for my needs. I don't have to change config often, maybe once a year or so.
    - battery423 6y ago
      " I don't see a reason to add more complexity to something that is working fine for my needs" valid point though. I would do it due to: - keeping 'meantime to recovery' low due to having my configs in git -> fast recovery - using something like ansible for your small case to have it in git (instead of a shellscript) -> backup - perhaps a jenkins pipeline with testing the config before applying it automatically -> for quality - automatisation for having fun with it, seeing what is out there besides ssh and manuel editing At the end of the day, i assume that this automatisated process is also less dependend from myself. So i'm not sure who is doing things when you are on holiday or sick. I'm now having a small k8s cluster at home. I have never had anything which allows me to have my whole setup in git so easy with autohealing, domain management and high availability.