1.3.0-beta - a file deleted from local folder after local rename #640
Comments
|
I cannot post the logfile because it is too large (600MB). Let me know how would you like to get the log. kuba |
|
I started a 1.3.0-beta2 client on my directory with the state left from the previous run with beta1 as described above. The problem persists. The logical error is still in the code: every iteration of the mirall update cycle the number of sqlite_step errors increases. So the issue is not solved in beta2 yet. |
|
I just tried to reproduce that with a stable5 server (upcoming 5.0.7) and the current 1.3.0 beta2 with some patches from today, which I don't consider changing anything. For me, local renaming of the file results in a MOVE command on the WebDAV server and works like a charm. @moscicki do you think you can isolate a testcase with one file or so and get me the client log file and if possible the access_log? Thanks. |
|
Hi, I think it may have to do with creating/modyfing the file via the web interface in the first place, before the file is moved locally and synced by the client. The client log is here: http://mail.kde.org/pipermail/owncloud/2013-May/009278.html The net result is that the file is missing on the client (is deleted), while on the server it exists. An excerpt of access_log of the apache server: 137.138.32.188 - octuser2 [23/May/2013:16:17:46 +0200] "GET /owncloud/status.php HTTP/1.1" 200 75 "-" "mirall-1.3.0beta1" kuba |
|
Can you try 1.3.0 nightly builds? |
|
We now released beta3. Can you test with that? |
|
Hi, yes, sure, thank you. I will try to come back to you tomorrow with a result. kuba On Jun 12, 2013, at 4:47 PM, Daniel Molkentin notifications@github.com wrote:
|
|
Hello, TEST1: I started beta3 client at the point where I left it from the previous run. The file is still not downloaded, however the log file looks differently now. See below. Observations:
So beta3 does not recover from that bad state. I think it is important that mirall is able to recover from bad state. TEST2: I deleted a .csync_journal.db on the client to see if anything changes. The file is not synced however there is a different error in the log file. Maybe this will be of some use for you. TEST3: TEST4: LOGFILES: TEST1: 06-14 17:57:34:558 #-------# oC found on "http://box.cern.ch/owncloud" TEST2: 06-14 18:31:51:929 oc_module: owncloud_stat owncloud://box.cern.ch/owncloud/remote.php/webdav/web1.txt called 06-14 18:31:52:342 _insert_metadata_visitor: SQL statement: INSERT INTO metadata_temp |
|
Here is my analysis from the logs: Test1: The file was blacklisted in the database because of previous errors. Which is why 'nothing' happens. Test2: Apparently the GET on owncloud://box.cern.ch/owncloud/remote.php/webdav/web1.txt returns with 404 not found. The part of the log that does the update is missing, but since the client knows about web1.txt, it is probably what is returned by the server. So maybe the server is laying and pretends it has a file which it does not have. Maybe some wrong cache on the server? |
|
any solution for this?? |
|
We removed the blacklist in 1.3 final. We still don't know wh the server had an error. |
|
Is this bug still valid? |
|
Yes its valid.. I just lost my vacation photos. Lucky for me I still have
|
|
I am working through ownCloud for an application. I have not encountered this problem. |
|
reverting the file names to its previous names brings back lost files. |
|
I retested this again with a 5.0.10 server. I was not able to reproduce an fail with a file or directory created in the web interface, synced to the client and rename it on the client machine. It worked all the time, also with proper MOVE commands sent to the server. I will close this report, please reopen or create a new bug if the problem persists. Thanks! |
Hello,
I setup a brand new test account both locally and on my owncloud server, both called octuser2. I created a file via the web interface, it got synced to a local client. Then I renamed the file locally and after a while it was deleted by mirall. However the file appears OK on the server.
Here is the detailed log:
test client: 1.30-beta compiled on slc6
test server: owncloud-5.0.3-17.1.noarch (note: after upgraded the server to the newest version (5.0.6-3.1) the issue persists)
/tmp/mirall-build/bin/owncloud --logfile ~/mirall.log &
created web1.txt via WEB INTERFACE - got synced to the client
[octuser2 ownCloud]$ sqlite3 -column -header .csync_journal.db "select * from metadata limit 10;"
phash pathlen path inode uid gid mode modtime type md5
2121407186260769061 8 web1.txt 6422578 9406 9406 33188 1369318729 0 519e2549973ae
[octuser2 ownCloud]$ ls -li
total 0
6422578 -rw-r--r--. 1 octuser2 octuser2 0 May 23 16:18 web1.txt
renamed the file:
[octuser2 ownCloud]$ mv web1.txt web1-rename.txt
[octuser2 ownCloud]$ sqlite3 -column -header .csync_journal.db "select * from metadata limit 10;"
phash pathlen path inode uid gid mode modtime type md5
4189687575036615272 15 web1-rename.txt 6422578 9406 9406 33188 1369318729 0 519e2549973ae
PROBLEM #1: change not reflected in the web interface even after page reload
On the server, however:
the file exists with a correct name:
ls -l /cloud_data1/octuser2/files/web1-rename.txt
the database table also seems to have a correct path:
mysql> select * from oc_filecache where path like "%web1%.txt";
+--------+---------+-----------------------+----------------------------------+--------+----------+----------+----------+------+------------+-----------+---------------+
| fileid | storage | path | path_hash | parent | name | mimetype | mimepart | size | mtime | encrypted | etag |
+--------+---------+-----------------------+----------------------------------+--------+----------+----------+----------+------+------------+-----------+---------------+
| 53171 | 22 | files/web1-rename.txt | cdfc9cf0f8cc9bfc83ffd6f175f7dc95 | 53163 | web1.txt | 10 | 5 | 0 | 1369318729 | 0 | 519e2549973ae |
+--------+---------+-----------------------+----------------------------------+--------+----------+----------+----------+------+------------+-----------+---------------+
PROBLEM #2:
after a while running of mirral the file disappeared from the client local folder and local state db altogether:
[octuser2 ownCloud]$ sqlite3 -column -header .csync_journal.db "select * from metadata limit 10;"
however it is still on the server:
[root]# ls -l /cloud_data1/octuser2/files/*
-rw-r--r--. 1 apache apache 0 May 23 16:18 /cloud_data1/octuser2/files/web1-rename.txt
-rw-r--r--. 1 apache apache 0 May 23 17:16 /cloud_data1/octuser2/files/web2.txt
and in servers db:
mysql> select * from oc_filecache where path like "%web1%.txt";
+--------+---------+-----------------------+----------------------------------+--------+----------+----------+----------+------+------------+-----------+---------------+
| fileid | storage | path | path_hash | parent | name | mimetype | mimepart | size | mtime | encrypted | etag |
+--------+---------+-----------------------+----------------------------------+--------+----------+----------+----------+------+------------+-----------+---------------+
| 53171 | 22 | files/web1-rename.txt | cdfc9cf0f8cc9bfc83ffd6f175f7dc95 | 53163 | web1.txt | 10 | 5 | 0 | 1369318729 | 0 | 519e2549973ae |
+--------+---------+-----------------------+----------------------------------+--------+----------+----------+----------+------+------------+-----------+---------------+
Below is the mirall logfile. It looks like a clear bug - the number of csync_statedb_inster sqlite_step errors grows in time for one file entry - I suspect this is the entry of the renamed file.
I have only two files in this directory.
The text was updated successfully, but these errors were encountered: