<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=utf-8">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<div dir="auto">Following bugs were reported against trash
translator in a brick multiplexing enabled environment:</div>
<div dir="auto"><a
href="https://bugzilla.redhat.com/show_bug.cgi?id=1447389"><br>
1447389</a></div>
<div dir="auto"><a
href="https://bugzilla.redhat.com/show_bug.cgi?id=1447390">1447390</a></div>
<div dir="auto"><a
href="https://bugzilla.redhat.com/show_bug.cgi?id=1447392">1447392</a></div>
<div dir="auto"><br>
</div>
<div dir="auto">In all the above cases trash directory, namely
.trashcan, was being created only on one brick per node.<br>
Trash directory is usually created within notify() inside trash
translator on receiving CHILD_UP event<br>
from posix translator. [trash.c:2367]. This CHILD_UP event is sent
by posix translator on receiving PARENT_UP.</div>
<div dir="auto"><br>
</div>
<div dir="auto">When brick multiplexing is enabled it seems that
notify() is invoked only on the first brick which follows normal<br>
path, but not in other bricks. On further debugging, we could see
that in glusterfs_graph_attach(), apart from <br>
graph preparation and initialization it lacks xlator_notify() and
parent_up calls as in normal cases.</div>
<div dir="auto"><br>
</div>
<div dir="auto">Can you please shed some light on how we can move
forward with these bugs?</div>
<div dir="auto"><br>
Also thanks Atin and Nithya for help in debugging above issues.<br>
<br>
</div>
<div dir="auto">Regards,<br>
Jiffin<br>
</div>
<div dir="auto"><br>
</div>
</body>
</html>