|Priority:||4 - Normal|
|Created by:||Patrick Mooney [X]|
|Reported by:||Patrick Mooney [X]|
|Assigned to:||Patrick Mooney [X]|
Fixed: A fix for this issue is checked into the tree and tested.
(Resolution Date: 2018-11-26T16:01:37.547Z)
2018-12-06 Grizz (Release Date: 2018-12-06)
Presently, if a client connected to one of the bhyve UART sockets disconnects, it will not be able to reestablish a meaningful connection until the instance restarts. A concrete case of this would be a restart of the metadata service while instances are running, preventing any of them from being able to perform mdata-* requests after that.
As part of the fixes for OS-7369, we should look to improve this edge case.
To test this, I started several bhyve VMs on a CN. After verifying that they could access the metadata service via ttyb/COM2, I restarted the metadata service (both via svcadm and
pkill -9). When the service had restarted, I checked to see that I could still successfully make metadata requests from inside the instances.
illumos-joyent commit bb7f525820fb9ee32d8b59ed1b4c53c0624d8598 (branch master, by Patrick Mooney)
OS-7369 full UART FIFO can send bhyve into a spin
OS-7381 bhyve should handle UART socket reconnect
Reviewed by: Hans Rosenfeld <firstname.lastname@example.org>
Reviewed by: Mike Gerdts <email@example.com>
Approved by: Mike Gerdts <firstname.lastname@example.org>