This never appeared. Further testing shows only two people are
sufficient to make an import stall
C:\src\svn-test>svn --version
svn, version 0.29.0 (r6976)
compiled Sep 8 2003, 00:45:48
A test was performed with several parallel imports. The imports
proceeded very slowly after the fourth importer was started and none
completed prior to database corruption.
Definition "very slowly" - so slowly that 30 minutes after the first
import loop noticed the database error shown below the other import
loops are still stalled and haven't noticed the database error yet.
The import loops did become progressively slower and each seemed to
have gotten about 2,200 files into the import.
The test data being imported consisted of a tree of 11,456 files and
directories of mixed text and binaries for a total of 204,443 KB.
Eight instances of the following loop were run against a newly created
repository, each targeting a different tree within the repository.
C:\src\svn-test>type t.bat
:loop
svn import -m - i:/james/b file:///src/svn-test/repos/%1
svn rm file:///src/svn-test/repos/%1
goto loop
Between five and eight hours after this was started on a
dual-processor NT5 workstation one loop started getting this error:
C:\src\svn-test>C:\src\svn-test>svn import -m - i:/james/b file:///src/svn-test/
repos/2
svn: Couldn't open a repository.
svn: Unable to open an ra_local session to URL
svn: Unable to open repository 'file:///src/svn-test/repos/2'
svn: Berkeley DB error
svn: Berkeley DB error while opening environment for filesystem /src/svn-test/re
pos/db:
DB_RUNRECOVERY: Fatal error, run database recovery
At no point did CPU usage get above 10%, nor did memory usage exceed about 25%
of physical RAM: there is no apparent resource issue at all.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Sep 20 23:44:58 2003