On Wed, Apr 24, 2019 at 02:37:36AM +0000, Zhangbo (Oscar) wrote: > >From: Stefan Hajnoczi [mailto:stefa...@gmail.com] > >Sent: Tuesday, April 23, 2019 9:57 PM > >On Tue, Apr 23, 2019 at 11:12:40AM +0800, Jie Wang wrote: > >> Does qemu-pr-helper have its own log files？ > > > >No, it does not. qemu-pr-helper uses stderr for errors and there is not > >much output during normal operation. > > > >Check journalctl(1) for errors if you are launching it as a systemd > >service. > > > >Stefan > > The error messages that sent to *stderr* are *not* captured by anyone, then > these error messages seems helpless.
This depends on how qemu-pr-helper was launched. Daniel Berrange mentioned that stderr is not captured when libvirt launches qemu-pr-helper. So in this case there is a problem. > It's different form qemu's log, qemu's stderr logs are transferred to > virtlogd, > that means these logs are finally written into disk. > Any further development plan? > > We thought of: > 1 logging into syslog, but the malicious guest may cause large amount of error > messages to overflow the message log. > 2 or, let virtlogd to redirect its logs. But if virtlogd get restarted, the > helper's log would > no longer be logged, because, unlike qemu, there's no qmp socket between > libvirtd and the helper, the channel between the helper and virtlogd could > not be > re-established. > 3 or, let qemu to log the helper's log, seems a little odd and complex. > 4 or, let the helper realize its own logging scheme and log into itself's > logfile ? Perhaps libvirt should be modified to handle stderr. Stefan
signature.asc
Description: PGP signature
