marking a host down doesn't mark it's devices down.

[GoogleCodeExporter]

What steps will reproduce the problem?
mark a host down & !watch on the console. 

What is the expected output? What do you see instead?
monitor jobs continue to poll the host for usage and replicate jobs continue to 
try and copy files from the devices.

What version of the product are you using? On what operating system?
v2.53 on ubuntu 2.6.32-32

Please provide any additional information below.

Original issue reported on code.google.com by jon.qu...@gmail.com on 4 Nov 2011 at 4:56

[GoogleCodeExporter]

Does it keep monitoring the devices forever, or does it stop after 15-30 
seconds? There's a short delay before host status is updated everywhere.

Original comment by dorma...@rydia.net on 4 Nov 2011 at 5:06

[GoogleCodeExporter]

Looks like it's always done this... not a new bug.

replication checks dev->dstat->can_read_from, but should use 
dev->can_read_from, which should check its host state as well as the device 
(similar to should_get_new_files).

The monitor will check everything that's reachable and doesn't check the host 
state.

Going to punt on a fix for a later version; I guess mark all your devices as 
down if you intend to take that host down for a bit.

Original comment by dorma...@rydia.net on 14 Nov 2011 at 10:52

[GoogleCodeExporter]

marking as accepted

Original comment by dorma...@rydia.net on 14 Nov 2011 at 10:52

• Changed state: Accepted

[GoogleCodeExporter]

Noting: after 2.50 the bug is consistently broken. marking a host down will 
mean even normal reads still happen.

now that all parts are consistent, it should be fixed in its entirely soon :P

Original comment by dorma...@rydia.net on 8 Mar 2012 at 8:11

[GoogleCodeExporter]

This got fixed way back in 2.64. closing!

Original comment by dorma...@rydia.net on 3 Feb 2013 at 12:17

• Changed state: Fixed