[Linux-cluster] Ricci doesn't work
jpokorny at redhat.com
Wed Sep 12 12:00:09 UTC 2012
On 10/09/12 21:54 +0000, Chip Burke wrote:
> Well, after a few days of testing it seems luci/ricci is still flakey.
At first, there seems to be more than "XML entities" problem involved.
This is because luci, unlike ccs and ccs_sync ("cman_tool version" uses
ccs_sync under the hood) should not suffer from that one.
So probably yet another separate issue...
> If I update things via luci, the cluster.conf on the local machine with
> luci updates, but nothing pushes out to other nodes. If I then manually run
> ccs_sync on the node with the new configuration, things push out to the
> other nodes fine. While this is wonky, at least it is consistent and
> repeatable so I can do things in this manner until fixes are in. Though,
> certainly let me know if you want any more logs or whatever from me.
To get a better error message in luci.log, something to start with, could
you apply a workaround for that "no translator" issue, please?
The recipe, based on the real patch to fix the mentioned bug, is as
follows (as root on the host with luci installed):
pushd $(rpm --eval %python_sitearch)
cat <<EOF | patch --fuzz=3 -b .std
diff --git a/luci/lib/ricci_helpers.py b/luci/lib/ricci_helpers.py
@@ -7,6 +7,9 @@
+from pylons.i18n.translation import _get_translator
from luci.lib.helpers import ugettext as _
from luci.model import DBSession
@@ -29,6 +33,14 @@ class PWorker(threading.Thread):
self.cluster_members_only = cluster_members_only
+ # see http://comments.gmane.org/gmane.comp.web.turbogears/46896
+ # this is stolen from the pylons test setup;
+ # it will make sure the gettext-stuff is working, that is
+ # we inject translator object to this private thread similarly
+ # as it is done by the framework in per-request threads
+ translator = _get_translator(None)
if len(self.triples) == 0:
@@ -67,6 +79,8 @@ class PWorker(threading.Thread):
self.ret[triple] = r
def send_batch_parallel(triples, max_threads, cluster_members_only=False):
mutex = threading.RLock()
threads = list()
If luci.log still does not provide any better information, please run this
sed one-liner (once the luci was started at least once so the respective
file is assuredly in place; luci restart needed afterwards):
sed -i.std "/logger_luci/bone;b;:one s|INFO|DEBUG|;ttwo;n;bone;:two n;btwo" \
More information about the Linux-cluster