[
https://issues.apache.org/jira/browse/COUCHDB-2821?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14902487#comment-14902487
]
ASF GitHub Bot commented on COUCHDB-2821:
-----------------------------------------
GitHub user rnewson opened a pull request:
https://github.com/apache/couchdb-fabric/pull/31
Don't create databases with toolong shard names
COUCHDB-2821
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/cloudant/couchdb-fabric 2821-dbname-length
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/couchdb-fabric/pull/31.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #31
----
commit 895d203aa49212aee192fba9efea4a985193ebd4
Author: Robert Newson <[email protected]>
Date: 2015-09-22T11:52:23Z
Don't attempt creation of a database that already exists
commit c22a727a790446dd37797bc130579dc4d8279911
Author: Robert Newson <[email protected]>
Date: 2015-09-22T12:10:23Z
Don't create dbs doc if shard files fail for enametoolong reason
COUCHDB-2821
----
> Prevent crash if user creates a database with a very long name
> --------------------------------------------------------------
>
> Key: COUCHDB-2821
> URL: https://issues.apache.org/jira/browse/COUCHDB-2821
> Project: CouchDB
> Issue Type: Improvement
> Security Level: public(Regular issues)
> Reporter: Robert Newson
>
> In couchdb before 2.0 a database create request would cleanly fail if the
> constructed filename for the database was too long for the underlying
> filesystem.
> In 2.0, the act of creating shard files and saving the shard map to the 'dbs'
> database happens concurrently. The latter will succeed even if the shard file
> creations fail, even if the reason for failure is fatal. CouchDB will also
> try to create these files if they are missing when queried. It is easy for
> the severity and frequency of those errors to take down the supervision tree.
> Guard against this by enhancing the database name validation function to
> reject names where any path segment exceeds 128 characters. Almost all
> filesystems can handle file and directory names of 255 characters at least
> but we constrain ourselves to half of this to allow room for system overhead
> (we already add a timestamp, for example).
> By 'path segment', I mean that 'foo/bar/baz' has three segments. When
> translated to disk, there will be a baz.couch file in the bar directory which
> is in the foo directory. We can be less aggressive than applying a maximum
> database name length, it is sufficient for each part to be under the limit.
> For exotic cases where names above 128 characters are desired, the use of
> multiple path segments could serve as a useful workaround.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
