I don't think its too much to ask people to create their ZK paths:
zkCli.sh -server localhost:2181 create /kafka kafka
Asking a ZK application to use a non-existent chroot is like asking tar
to extract to a non-existent directory. An error is appropriate. I think
the fix here is to give a better error than the very unhelpful/cryptic
"Path length must be > 0" (https://issues.apache.org/jira/browse/KAFKA-294)
Cheers
-David
On 11/13/13 11:13 PM, marek wrote:
Thank you Joel, got it.
I cloned the git repo and am working in trunk.
This is the correct branch for the latest in 0.8, is that right?
Or should I work against one of the other branches?
-Marek
On Wednesday, November 13, 2013, Joel Koshy wrote:
Marek was referring to the zookeeper namespace - Marek, there are
jira(s)
https://issues.apache.org/jira/browse/KAFKA-294
https://issues.apache.org/jira/browse/KAFKA-404
Should be straightforward to finish that up - want to give that a go?
Thanks,
Joel
On Thu, Nov 14, 2013 at 03:48:01AM +0400, Kostya Golikov wrote:
Just one word: Windows.
I'm not a maintainer, but AFAIK kafka targets Windows as well and isn't
chroot is a Unix-specific feature?
2013/11/14 marek <mdma...@gmail.com <javascript:;>>
Hey,
I'm trying to feel-out if auto creating the broker's registration path
on server start up would be wanted as a code contribution.
How do the maintainers feel about:
1.
Having configurable server parameter:
auto.create.chroot = true | false
2.
If (1) is set to true, the chroot would be create automatically in the
case that the it is missing from zookeeper, and the server would start
up normal.
If the maintainers are OK with this idea I'll code it up, it would
solve a use case we have at my work, and I would send in a patch or
follow any contribution process given.
Right now if you specify zookeeper.connect as:
host1:port1,host2:port2,host3:port3/chroot/path
and /chroot/path does not exist in zookeeper, the server does not
operate.
-Marek
