reveals that rhgb spins at 100% cpu during boot. Some inspection
of the source code reveals this is do to the handling of
rhgb-console... it calls poll() on a fifo ... and fifo's have
the behavior that once the last writer closes the fifo, it
reports HUP until a new reader opens it.
Googling on the topic reveals:
with an IRC snippet from myself about the topic.
Two possible workarounds are suggested there:
- Close the fifo and reopen it ...doesn't work here since
we might lose data
- open the fifo read/write so there are never 0 writers.
Basically, my conclusion has always been that fifo's are just
evil, poorly defined creatures. I think the O_RDWR hack will
work, at least on linux. I'm not sure it wouldn't be long
term to replace '> /etc/rhgb/temp/rhgb-console' with
'| rhgb-client --echo' or something.
I see, ouch ... I will try to fix ASAP.
I assume this also affects RHEL 3 U2 and beyond since rhgb is now
included. Is that correct? Will an update to rhgb be available
RHEL 3 U2 seems to use rhgb-0.11.2 which is not affected by that bug.
Updating to the latest doesn't sound resonnable it would require to
upgrade a number of other pieces.
The original bug has been fixed with a FC3 update rhgb-0.15.1