Bugzilla – Full Text Bug Listing
| Summary: | Subversion repository for the technology projects does not work | ||
|---|---|---|---|
| Product: | Community | Reporter: | Alexander Gurov <a.gurov> |
| Component: | Subversion | Assignee: | Eclipse Webmaster <webmaster> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | blocker | ||
| Priority: | P3 | CC: | andreas.buchen, aog-ecl, elena.nayashkova, karl.matthias, mathias.stuempert, mknauer |
| Version: | unspecified | ||
| Target Milestone: | --- | ||
| Hardware: | All | ||
| OS: | All | ||
| Whiteboard: | |||
|
Description
Alexander Gurov
This weekend we attempted to split the monolithic Technology repository into smaller per project repositories. However this failed and has left the repository in an unstable state, as such I have set the repository as read only and have asked for the data to be reloaded from fridays archive tape. -M. Could you please specify a date (approximately) when reloading of the repository could be done? As I said it's in progress, so hopefully this will be finished in a few(~3) hours. -M. The restore is complete and everything looks good. Please let me know if there are any problems. -M. I've gotten a couple of reports of missing commits. Re-opening while I investigate. -M. Matt, just a quick notice from the org.eclipse.mat repository: My Friday evening commit is included, but some older commits are missing. I just wanted to let you know, because this doesn't fit to the "updates between Friday's backup and the time when the move started" assumption. Some additional information: The latest revision number which is saved in my working copy is 20401. Now the latest revision on the repository is 20094. So, at least 307 revisions are lost on whole repository. Thanks for that info Alex, I appreciate it. And Andreas, I'm surprised that there are missing older commits, but I'll see what there is to be seen. At this time the restore is still running, once that's done I'll load it up beside the current location and take a look at the differences. Once I think things should be ok I'll touch base via email to have you guys check things out and see what you see as missing/odd. -M. Matt, you are right. I got confused by the error messages by the API tooling. The build of August 8th has become our new "internal" baseline. The I couldn't build, this led me to believe I have an inconsistent snapshot of the repository. But, alas, this is not the case. I hope you can recover the old commits. Andreas. *** Bug 243985 has been marked as a duplicate of this bug. *** Hello Matt, Any ideas when repository could be alive? An update: I've talked to our ISP and they tell me that they are having difficulties loading one of the tapes in the backup set, but they are working on it. -M. OK, we understand that the task is pretty complex. But could ISP tell you also not exact, but approximate date when the repository could be available? We need some information to correct our plans. A further update. I received this message from our ISP last night at 6:30: 'Just to give you an update on the situation. The system administrators are in contact with technical support for our managed backup system. It seems that the server will not take the tapes that are needed at this time. They have trouble ticket open with them that is, of course, marked urgent.' However I have been getting in touch with the project leads to have them check out the 'aborted' svn move area, to see if I had made an overly cautious assessment that everything had gone wrong. I'm getting a few responses back that indicate that for those projects the aborted move data is matching their local working copies. My plan at this time is this: I'm going to re-archive the current out of date technology repo, and as I hear from the projects I'll do one of two things: 1)If the aborted move matches their local copies, I'll move that aborted repository into /svnroot/technology, setup the Http/https access, and flag it as read/write for the project to begin working. 2)If the aborted move doesn't match their local copy, then they and I will have to wait for the restore to complete. Now that isn't entirely true, I'm going to see if it's possible for me to 'merge' SVN dumps so I build an amalgamated dump file for those projects to get them on their way. That's where we stand this morning. -M. Another update from our ISP: 'Unfortunately, the restore is still taking place. No ETA was given at this time. They have however fixed the tape issue late last night. I will keep you posted.' So my optimistic guess is that I should have the data back Friday morning at which point I'll work with the projects that have not confirmed the aborted move data is ok to get them back up and running. -M. I just received the restore data. I'm going to be confirming the move procedure for the outstanding projects at this time and I want to have the move finished tonight. -M. Ok, the org.eclipse.subversive repository has been put back in place and put in read/write mode. The restored data was short 20 commits based on the revision count for last week, but I don't know which project(s) are affected. At any rate you'll have to do a check to see if it's your project and then just re-commit the changes that exist locally but not at Eclipse.org. I'm still working on org.eclipse.higgins as they are the last outstanding project. However flattening their repository is proving problematic as it's involving a lot of manual edits to the dump file. -M Alright I've finally gotten the Higgins project back. With that I'm closing this bugs as every technology project should now be able to commit code into their shiny new separated repositories. -M. Thanks for your perseverance, Matt. I've seen you walk into the office looking as miserable as a boxer who has been defeated in round 15, so the SVN tools for administrators are definitely lacking. 2 thumbs up for Matt! 2 thumbs down for SVN tooling! Not the first problem it has caused... Movng to Subversion component. |