Open Stack

Shyam,

The proxy server log entry:

Jul  4 10:08:32 ip-10-141-173-100 proxy-server: Object PUT returning 503 for [408, 201] (txn: txa3cac3a103d54f4d84bed-0053b67a56) (client_ip: 127.0.0.1)

Indicates that one of the object server PUTs returned a status of 408, which is a request timeout while the other object server  PUT returned 201 i.e. success.  You need to figure out why one of the PUTs failed!  Take a look at transaction txa3cac3a103d54f4d84bed-0053b67a56 in the object server logs, it *may* help identify why one PUT failed.

Gerry.

From: Shyam Prasad N [mailto:nspmangalore at gmail.com]
Sent: 04 July 2014 11:53
To: Drudy, Gerry
Cc: openstack at lists.openstack.org
Subject: Re: [Openstack] Replication count..

Hi Gerry,
Thanks for the reply.
Attached the outputs for the commands you asked for.
I see a python traceback while reading container metadata. But I've configured replica count as 2 for container rings too (all 3 types of rings have been similarly configured). And it's able to read the other replica.

I wonder why this error then?

On Fri, Jul 4, 2014 at 4:00 PM, Drudy, Gerry <Gerry.Drudy at hp.com<mailto:Gerry.Drudy at hp.com>> wrote:
I too would have expected the PUT to succeed given the config you describe.  Perhaps the proxy log entries for the PUT might indicate why it is failing?

If you can’t figure out what is going wrong from the proxy server logs can you post the o/p of the following 2 commands:

swift-ring-builder /etc/swift/object.builder
swift-get-nodes /etc/swift/object.ring.gz blah blah blah

Gerry.

From: Shyam Prasad N [mailto:nspmangalore at gmail.com<mailto:nspmangalore at gmail.com>]
Sent: 04 July 2014 10:45
To: openstack at lists.openstack.org<mailto:openstack at lists.openstack.org>
Subject: [Openstack] Replication count..

Hi,
I have a swift setup with 2 disks each on 2 nodes. The swift rings have been setup with replication count of 2, with each node with unique zone id. After creation of a few objects, I pulled out one of the disks from node1. After some more object creations, I pulled out the remaining swift disk on node1. Now, my GETs are successful. But PUT requests are failing with "Service Unavailable" error.
I'm confused with this behaviour. I thought swift tries to PUT replicas for an object into different zones on a best effort basis. Right now, there are still two disks active on the other zone. Shouldn't it put both the replicas on the available disks now, rather than returning the above error? Or is my understanding about this wrong?
Thanks in advance for the replies.

--
-Shyam



--
-Shyam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20140704/4ec4aa4d/attachment.html>