If a deploy fails, log output to file for inspection #473

New Issue

2025-01-07T14:07:59Z

fauno commented

2025-01-07 14:07:59 +00:00

When an app fails to converge, you're suggested to run abra app logs but this only shows the latest lines, which don't necessarily indicate what happened. Is it possible to show the logs during deploy, on the same output? That way I could see what failed during deploy, instead of guessing when are logs available (I can't, always seem to get less output that I need)

When an app fails to converge, you're suggested to run `abra app logs` but this only shows the latest lines, which don't necessarily indicate what happened. Is it possible to show the logs during deploy, on the same output? That way I could see what failed during deploy, instead of guessing when are logs available (I can't, always seem to get less output that I need)

decentral1se commented

2025-01-07 16:18:00 +00:00

@fauno that's a very interesting idea 🤔 If others support this idea, we could think about wiring up a prototype to test it out. I think it would in theory give more clarity to what is going on?

And to check we're on the same page, if the deployment passes, the logs stop tailing? And if it fails, it stops tailing. And you get the same message at the end as per usual?

The current work-around is something like this after the deployment:

abra app logs <domain> \
  --since "$(date --date='10 minutes ago' -u +'%FT%T.%2NZ')"

~~But I now notice that this actually hangs with 0 output 😭 That's a 🐛~~

update: typo, ~~10 minutes~~ 👉 10 minutes ago works 🎉

@fauno that's a very interesting idea 🤔 If others support this idea, we could think about wiring up a prototype to test it out. I think it would in theory give more clarity to what is going on? And to check we're on the same page, if the deployment passes, the logs stop tailing? And if it fails, it stops tailing. And you get the same message at the end as per usual? The current work-around is something like this after the deployment: ``` abra app logs <domain> \ --since "$(date --date='10 minutes ago' -u +'%FT%T.%2NZ')" ``` ~~But I now notice that this actually hangs with 0 output 😭 That's a 🐛~~ **update**: typo, ~~`10 minutes`~~ 👉 `10 minutes ago` works 🎉

decentral1se added the enhancement label 2025-01-07 16:31:30 +00:00

fauno commented

2025-01-07 16:49:17 +00:00

Yeah that should do it! I didn't notice the --since flag :/

Yeah that should do it! I didn't notice the `--since` flag :/

decentral1se commented

2025-01-07 16:59:50 +00:00

@fauno cool! actually, it does work 😅 I just had a typo in the obscure date syntax in my comment above. hopefully that helps for now. but yeh, if you could try gather more input on this feature request, it'd be great!

@fauno cool! actually, it does work 😅 I just had a typo in the obscure `date` syntax in my comment above. hopefully that helps for now. but yeh, if you could try gather more input on this feature request, it'd be great!

decentral1se commented

2025-01-09 11:11:12 +00:00

#478 feels related.

https://git.coopcloud.tech/toolshed/abra/issues/478 feels related.

decentral1se referenced this issue

2025-01-11 11:02:48 +00:00

Failed update silently rolls back #478

decentral1se commented

2025-01-11 11:09:37 +00:00

@fauno any ideas how a layout might look in combination with #478 (comment)? (🤯) Ideally we'd have both the status updates from Docker itself and the logs tailing alongside somewhere. We don't have so much screen real estate but I'd love to combine. This could make things finally visible and informative... see #473 (comment)

~~@fauno any ideas how a layout might look in combination with https://git.coopcloud.tech/toolshed/abra/issues/478#issuecomment-22548? (🤯) Ideally we'd have both the status updates from Docker itself and the logs tailing alongside somewhere. We don't have so much screen real estate but I'd love to combine. This could make things finally visible and informative...~~ see https://git.coopcloud.tech/toolshed/abra/issues/473#issuecomment-22579

decentral1se added this to the abra v0.10.0 project 2025-01-11 11:09:45 +00:00

decentral1se commented

2025-01-13 09:13:08 +00:00

@fauno a wild idea appears: we can record the start time of the deployment. then, run the deployment. when #478 is implemented we will know for sure if a deployment failed or not. if a deployment fails, we can dump all the logs from all containers since the deployment start time directly without prompt (because you would do this anyway). if the deployment succeeds, we show no logs. how does this sound for a design?

@fauno a wild idea appears: we can record the start time of the deployment. then, run the deployment. when https://git.coopcloud.tech/toolshed/abra/issues/478 is implemented we will know for sure if a deployment failed or not. if a deployment fails, we can dump all the logs from all containers since the deployment start time directly without prompt (because you would do this anyway). if the deployment succeeds, we show no logs. how does this sound for a design?

👍 2

fauno commented

2025-01-13 12:44:56 +00:00

sounds great!

🎉 1

decentral1se commented

2025-01-15 09:19:30 +00:00

Alternatively, taking some parts of what @forest was talking about here we could also log all output of a failed deployment to file in ~/.abra/logs/<server>/<app>_<timestamp> (or something like this) for further inspection (and also ease of scp ... to fellow workers for inspection!

I think that also might be an easier and a more sustainable option, giving people the full overview but then instead of a potentially confusing dump of some limited amount (or worse, too much) of logs in the terminal all at once!

Alternatively, taking some parts of what `@forest` was talking about [here](https://matrix.to/#/!DfXPgKLoYCvjHithgS:autonomic.zone/$sO_fPQUrXwfBWtvDZH7TUjrk0dxCzVsKXn5e5Lo8a_Q?via=autonomic.zone&via=matrix.org&via=cyberia.club) we could also log all output of a *failed deployment* to file in `~/.abra/logs/<server>/<app>_<timestamp>` (or something like this) for further inspection (and also ease of `scp ...` to fellow workers for inspection! I think that also might be an easier and a more sustainable option, giving people the full overview but then instead of a potentially confusing dump of some limited amount (or worse, too much) of logs in the terminal all at once!

👍 2

decentral1se added the design label 2025-01-16 07:46:39 +00:00

decentral1se added the release-candidate label 2025-02-12 14:56:58 +00:00

decentral1se changed title from ~~Add logs to deploy command?~~ to If a deploy fails, log output to file for inspection

2025-03-13 09:51:14 +00:00

decentral1se removed the design label 2025-03-13 09:51:20 +00:00

decentral1se referenced this issue

2025-03-13 22:06:53 +00:00

Failed update silently rolls back #478

decentral1se self-assigned this 2025-03-16 11:48:10 +00:00

decentral1se referenced this issue

2025-03-22 12:28:59 +00:00