Attachment 691632 Details for Bug 680256 – Notes about patch behavior (should have been added a long time ago)

Notes about patch behavior (should have been added a long time ago)

680256.txt (text/plain), 2.04 KB, created by Lon Hohberger on 2013-02-01 15:57:07 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Lon Hohberger

Created: 2013-02-01 15:57:07 UTC

Size: 2.04 KB

patch

obsolete

>2-node cluster, with 2 services:
>
><service name="important" exclusive="1" max_restarts="1" restart_expire_time="3600" />
><service name="whocares" exclusive="2" />
>
>Suppose "important" fails and needs to be restarted; this happens 
>without the entire node failing - just some resource(s) in the service
>have stopped behaving reliably.
>
>
>Before 680256:
>
> + node1 ----+     + node2 ----+
> | important |     | whocares  |   (important failed; needs restart)
> +-----------+     +-----------+
>
> + node1 ----+     + node2 ----+
> |           |     | whocares  |
> +-----------+     +-----------+
>
> + node1 ----+     + node2 ----+
> | important |     | whocares  |   (uh oh, important failed, needs
> +-----------+     +-----------+    restart -again-, but this exceeds
>                                    threshold)
> + node1 ----+     + node2 ----+
> |           |     | whocares  |
> +-----------+     +-----------+
>
> [end] Lower priority service running, higher priority service offline
>
>After 680256:
>
> + node1 ----+     + node2 ----+
> | important |     | whocares  |   (important failed; needs restart)
> +-----------+     +-----------+
>
> + node1 ----+     + node2 ----+
> |           |     | whocares  |
> +-----------+     +-----------+
>
> + node1 ----+     + node2 ----+
> | important |     | whocares  |   (uh oh, important failed, needs
> +-----------+     +-----------+    restart -again-, but this exceeds
>                                    threshold)
> + node1 ----+     + node2 ----+
> |           |     | whocares  |
> +-----------+     +-----------+
>
> + node1 ----+     + node2 ----+
> |           |     |           |   (Lower prio service stopped to make
> +-----------+     +-----------+    space for higher prio service)
> 
> + node1 ----+     + node2 ----+
> |           |     | important |   (Higher priority service started
> +-----------+     +-----------+    on node 2)
>
> + node1 ----+     + node2 ----+
> | whocares  |     | important |   (lower prio service restarted on 
> +-----------+     +-----------+    empty node 1)
>
> [end] Both services running; they have switched nodes systematically
>
>
>

Actions: View

Attachments on bug 680256: 480852 | 480876 | 486712 | 691632