Empty or non-existent '/db/txn-current' in a Subversion repository
-
13-07-2021 - |
Question
I've encountered an issue with Apache Subversion repository and I'm looking for the best solution:
I can't commit to a Subversion repository and I'm getting the error:
svn: E720002: Commit failed (details follow): svn: E720002: Can't open file 'C:\Repositories\myproject\db\txn-current': The system cannot find the file specified.
If I create an empty txn-current
file if will get another error on commit:
svn: E070014: Commit failed (details follow): svn: E070014: Can't read file 'C:\Repositories\myproject\db\txn-current': End of file found
NOTE: Let's imagine that I don't have a backup of a repository so I have to fix the issue manually.
The same behavior has been already discussed on StackOverflow: svn commit problem, Cannot read 'd:\UC\db\txn-current': End of file found. However I doubt that answers there are really helpful:
I don't think that creating the
\db\txn-current
manually and putting a0
there (or any other integer) is a good solution. I suppose that it could result into some unexpected behavior in near future,svnadmin recover
and/orsvnadmin verify
do not create and/or populatetxn-current
. These commands are helpless in this case.
According to the FSFS description \db\txn-current
file contains the next transaction key number. If I put some integer to the file then I can commit and the integer will increase by 1.
However looks like putting some irrelevant integer can break something in my repo and I don't want this to happen.
The "txn-current" file is a file with a single line of text that contains only a base-36 number. The current value will be used in the next transaction name, along with the revision number the transaction is based on. This sequence number ensures that transaction names are not reused, even if the transaction is aborted and a new transaction based on the same revision is begun. The only operation that FSFS performs on this file is "get and increment"; the "txn-current-lock" file is locked during this operation.
What should I do to fix the \db\txn-current
issue properly? What integer should I put into the file? Is there any other way to solve this?
EDIT:
- How can I solve the same issue (non-existent or empty) but for
db/current
file?
Solution
In FSFS txn-current
holds next transaction id. Transaction ID stored in revision itself after committing. So puttign really big number in txn-current
should be fine.
Another solution would be to svnadmin dump
repository and then svnadmin load
it to fresh repository.
UPDATE: The issue with txn-current
file corruption was investigated and in fixed in r1483781.
- fix repository corruption on power/disk failure on Windows (r1483781)
Fix is available in Subversion 1.7.10 and later.
OTHER TIPS
I had similar problem. For resolving, I edited 'txn-current' file, put some large number say xxxxx in it. Along with it, I had to edit another file named 'current' in 'db\transactions' folder. I removed all contents from this file and just added a single enter and closed this file.
I was getting error as "Can't read 'D:\SVNROOT\UC\db\txn-current'.: End of file found" On probing further we realised that "Current" file in db folder was empty after some server OS upgrade activity. We found the latest revision number from revs folder and added the number into "current" file
It still did not work then we found this article which specifies that you have to put the revision number and then a "enter" after that
This works to resolve the error!
I had a similar problem, where a power failure occurred during a commit. I tried modifying the "Current" file but it did not work.
I made a backup of the repository, and then deleted the "current" file, and ran a "svnadmin verify" on the repository. Thereafter, I got the following error:
"Malformed file svn: Serialized hash missing terminator"
I then did a dump of the repository, deleted the old repository, recreated the Repository and imported the dump file.
Thereafter all worked well again. The full repository and History was still in place.
I did:
svnadmin verify /mnt/repos/
svnadmin recover /mnt/repos/
still got error...
svn: E000002: Übertragen schlug fehl (Details folgen):
svn: E000002: Kann Datei »/mnt/repos/db/txn-current« nicht öffnen: Datei oder Verzeichnis nicht gefunden
Then I created empty file: /mnt/repos/db/txn-current
then svn ci
worked again
If you don't want to wait:
svnadmin verify -r 997:HEAD <path>
is faster. exchange 997 with revision number, smaller than the bad revision
In my case. The VirtualMachine in which the SVN was setup, space was full. So cleaned up the space