From: <bac...@li...> - 2006-10-15 08:02:46
|
A BUGNOTE has been added to this bug. ====================================================================== http://bugs.bacula.org/bug_view_advanced_page.php?bug_id=0000687 ====================================================================== Reported By: Michael Brennen Assigned To: ====================================================================== Project: bacula Bug ID: 687 Category: Storage Daemon Reproducibility: always Severity: major Priority: normal Status: feedback ====================================================================== Date Submitted: 10-05-2006 21:42 PDT Last Modified: 10-15-2006 01:02 PDT ====================================================================== Summary: Multi-drive bug Description: In CVS for the past couple of weeks, multiple drives for multiple pools are not used; only one drive ends up being used. The two jobs using different pools are staggered by several hours, so one is always through before the other starts. I was finally able to reproduce what I think is the same problem with a regress test, two-pool-tape. The attached file was run at the default debug=100. If you want another run set higher, or a different test run, let me know. Now that regress is set up it should be easy to do. ====================================================================== ---------------------------------------------------------------------- kern - 10-07-2006 03:34 PDT ---------------------------------------------------------------------- Download attachment, then re-uploaded it without the underscore in the name which confuses Mantis (groan). Deleted the previous file. ---------------------------------------------------------------------- kern - 10-07-2006 03:42 PDT ---------------------------------------------------------------------- Unfortunately, your bug report is not well structured. The attachment that you uploaded contains a whole lot of stuff, and the instructions "Run the regress two-pool-tape test with the configuration in the attached file." are not valid. There are three configuration files for each regression script, so you must clearly indicate which one needs to be substituted, in addition, your script contains paths such as /bacula/... which do not exist on my machine. Please do the following: - Install an unchanged regression setup. - Configure it to run on your machine - Test a script - Modify one or more of the script/xxxx.conf.in files. - Run "make sed" - Re-run the test script to produce the error. - Upload the output produced from the script with debug 1 - Explain *clearly* what is not going right - Upload each of the script/xxxx.conf.in files you modified. ---------------------------------------------------------------------- Michael Brennen - 10-07-2006 10:33 PDT ---------------------------------------------------------------------- Sorry about the agglomeration of files into one; it occurred to me afterward that I had not even put in the original source file names to identify each one. Last week was rough and I did not do well at thinking through the details. I created the one file because I saw that mantis only had place for one upload; only after posting the bug did I understand that I could upload multiples one at a time after the fact. Thanks again... ---------------------------------------------------------------------- Michael Brennen - 10-07-2006 12:30 PDT ---------------------------------------------------------------------- The two-pool-tape test just failed again. Three new files are attached. In log1.txt there is this error: 07-Oct 13:49 localhost-sd: 3304 Issuing autochanger "load slot 1, drive 0" command. 07-Oct 13:54 localhost-sd: NightlySave2.2006-10-07_13.37.14 Fatal error: 3992 Bad autochanger "load slot 1, drive 0": ERR=Child died from signal 15: Termination. 07-Oct 13:54 localhost-fd: NightlySave2.2006-10-07_13.37.14 Fatal error: job.c:1732 Bad response to Append Data command. Wanted 3000 OK data, got 3903 Error append data Unfortunately I do not know how to tell you how to reproduce the error. When I run the test against actual hardware it fails. When I run the 2drive-incremental-2disk simulator the test passes; against my hardware it fails. If you can give some pointers where to look I will help where I can. ---------------------------------------------------------------------- kern - 10-13-2006 02:09 PDT ---------------------------------------------------------------------- I believe this should now be fixed in the CVS. See my comments for bug http://bugs.bacula.org/bug_view_advanced_page.php?bug_id=0000689. ---------------------------------------------------------------------- kern - 10-15-2006 01:02 PDT ---------------------------------------------------------------------- I suspect that your problem here is the same as bug http://bugs.bacula.org/bug_view_advanced_page.php?bug_id=0000689 -- i.e. you are using a broken mtx-changer script. Please correct that problem, re-run your tests and let me know if there is still a problem. Bug History Date Modified Username Field Change ====================================================================== 10-05-06 21:42 Michael BrennenNew Bug 10-05-06 21:43 Michael BrennenFile Added: multidrive_fail.txt 10-07-06 03:33 kern File Deleted: multidrive_fail.txt 10-07-06 03:33 kern File Added: multidrive.txt 10-07-06 03:34 kern Bugnote Added: 0001925 10-07-06 03:42 kern Bugnote Added: 0001926 10-07-06 03:42 kern Status new => feedback 10-07-06 10:33 Michael BrennenBugnote Added: 0001929 10-07-06 12:22 Michael BrennenFile Added: bconcmds.txt 10-07-06 12:22 Michael BrennenFile Added: log1.txt 10-07-06 12:22 Michael BrennenFile Added: log2.txt 10-07-06 12:30 Michael BrennenBugnote Added: 0001931 10-13-06 02:09 kern Bugnote Added: 0001949 10-15-06 01:02 kern Bugnote Added: 0001960 ====================================================================== |