openstack team mailing list archive
-
openstack team
-
Mailing list archive
-
Message #18282
Re: Something really weird when PUT overwrites a file with same name but different content 1.4.8 - BUG ?
A little update, this is happening updating headers too on existing objects
... were wondering whats wrong in our setup (a proxy ?) :S
An example setting a key-value on objects metadata (happens with any header)
curl -X HEAD -H "X-Auth-Token: ea5e7b70d58f4d45b0fc921e99b22f76" "
http://172.16.177.253:8080/v1/AUTH_1bf1f1b69a864abb84ed8a1bc82cff21/test-max/test";
-IHTTP/1.1 200 OK
*X-Object-Meta-Lean: CABEZA*
Last-Modified: Mon, 05 Nov 2012 14:59:04 GMT
Etag: d41d8cd98f00b204e9800998ecf8427e
Accept-Ranges: bytes
Content-Length: 0
Content-Type: application/octet-stream
X-Trans-Id: tx22818145f72b4ca3bf5f212721fecefc
Date: Mon, 05 Nov 2012 14:57:40 GMT
mvenesio@maxbox:~# curl -X PUT -H "X-Auth-Token:
ea5e7b70d58f4d45b0fc921e99b22f76" "
http://172.16.177.253:8080/v1/AUTH_1bf1f1b69a864abb84ed8a1bc82cff21/test-max/test";
-H "*X-Object-Meta-lean: PETERO" *-H "Content-Length: 0"
<html>
<head>
<title>201 Created</title>
</head>
<body>
<h1>201 Created</h1>
<br /><br />
</body>
</html>mvenesio@maxbox:~#
mvenesio@maxbox:~#
mvenesio@maxbox:~# curl -X HEAD -H "X-Auth-Token:
ea5e7b70d58f4d45b0fc921e99b22f76" "
http://172.16.177.253:8080/v1/AUTH_1bf1f1b69a864abb84ed8a1bc82cff21/test-max/test";
-IHTTP/1.1 200 OK
*X-Object-Meta-Lean: CABEZA <<---- OLD HEADER :S*
Last-Modified: Mon, 05 Nov 2012 14:59:04 GMT
Etag: d41d8cd98f00b204e9800998ecf8427e
Accept-Ranges: bytes
Content-Length: 0
Content-Type: application/octet-stream
X-Trans-Id: txb2186e54462b473fabaf0342e0339a05
Date: Mon, 05 Nov 2012 14:57:59 GMT
On Mon, Nov 5, 2012 at 10:21 AM, Leandro Reox <leandro.reox@xxxxxxxxx>wrote:
> Hi guys,
>
> We're facing a really weird random issue on Swift 1.4.8, when we update an
> object that already exists with the same name, but different content,
> sometimes randomly the operation, dispites the proxy answer ( always 201
> created) doesnt succeed, and the object content doesnt get updated. But
> sometimes the operation succeed and the object content get overwritten.
>
> What we saw in terms of replication :
>
> - When the object doesnt get updated we dont see the .db file replicated
> accross the cluster
> - Whe the object overwrites OK we saw the 3 replicas of the object and the
> 3 db replicas propagating OK accross the hole cluster.
>
> This is happening with all the containers, and all the objects, randomly
>
> Things we already checked:
>
> - Ring files replicated ok acrros all proxies and all datanodes
> - Confs files ok
> - Memcached partners ok
> - Replication services, and all the other services running ok on all nodes
>
> Is this really and issue or we should DELETE and object previous uploading
> with the same name ?. We found this really weird (sounds like a bugs) cause
> succeeds randomly
>
> Best !
> Lean
>
Follow ups
References
