Opened 14 years ago
Closed 13 years ago
#158 closed enhancement (invalid)
send guest consoles to zephyr
| Reported by: | geofft | Owned by: | |
|---|---|---|---|
| Priority: | normal | Milestone: | |
| Component: | internals | Keywords: | |
| Cc: |
Description
Once we resolve #157, it would be cute if we can get guest console output (perhaps starting once the guests have booted up and presented a logini prompt) to get sent to zephyr, so we see things like oopses and OOM kills and panics and other stuff that's indicative of a problem but unlikely to get logged from within the host and successfully reach zephyr that way.
Change History (3)
comment:1 follow-up: ↓ 2 Changed 14 years ago by mitchb
comment:2 in reply to: ↑ 1 Changed 14 years ago by geofft
Replying to mitchb:
I think we should reject this idea. It seems like a security problem to me to do something that might cause console output from an interactive session to get sent over zephyr...
Oh, point. Can... we have two console devices or something awful like that? (Both would need to get kernel messages.)
All of the guest console output is logged by conserver to a file on the host... I don't think I've ever been aware of that losing a crash message, have you? Which part of the need here is that not satisfying?
No, I haven't been aware of that losing a crash message, but it's not a notification, just a log. I often find servers acting funny, connect to their log, and go "Oh look an oops". It would have been nicer to know about it when the oops happened instead of when the problems become visible at the application level.
comment:3 Changed 13 years ago by achernya
- Resolution set to invalid
- Status changed from new to closed
-> scripts / trac-158 / mitchb 18:51 (recursive mutex enter? fsck that!)
This is moot now. We don't need anything convoluted like that. We have a dead kernel checker. And no, we already have rules to get the relevant stuff we want over zephyr. The hosts don't do that yet.
I think we should reject this idea. It seems like a security problem to me to do something that might cause console output from an interactive session to get sent over zephyr... we already have a PAM issue that needs fixing which might cause some sensitive info to get zephyred, and shouldn't be adding more.
All of the guest console output is logged by conserver to a file on the host... I don't think I've ever been aware of that losing a crash message, have you? Which part of the need here is that not satisfying?