git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: sadly requesting help

From
Samuel Tardieu <sam@rfc1149.net>
Date
Mar 31, 2010, 08:53 UTC
Message-ID
<2010-03-31-10-53-47+trackit+sam@rfc1149.net>
In-Reply-To
<eaacf13c1003301543r45bb18b2n1d10d209f01e3326@mail.gmail.com>
>>>>> "Raymond" == Raymond Auge <raymond.auge@liferay.com> writes:

Raymond> We had a colo failure over the last day or so and lost the last Raymond> 50-60 commits on our subversion repository (apparently, our Raymond> backup strategy was not granular enough).

Raymond> Luckily I use Git locally using the git-svn conduit.

Raymond> I need to rewind my repository to an earlier revision and I'm Raymond> hoping not to have to rebuild my local repo as the project is Raymond> huge and takes me at least 16 hours to checkout using git-svn.

I know this does not answer your question, but why not use this incident to switch to git, or at least to seriously investigate a possible future switch to git?

As you probably know, with git it would have been really easy to restore the full repository if at least one person does have a local copy of each branch (typically, the last person to have committed on a branch is likely to still have a full copy of the branch). And backups can be done simply by running "git fetch" from a secondary machine at regular intervals.

Success story: the company I worked for in 2008 had a similar incident a few months after we switched from svn to git. Not only were we able to restore a full repository copy, but also we were able to work in the meantime by setting one of the developers machine as the central repository, and development work was not disrupted for more than one hour (we had to educate some developers who were not familiar with setting remotes other than "origin" and pushing to them). We ran this degraded setting for a few days (degraded because we lost continuous testing and packaging capabilities that ran on the main server, and developers had to run the test themselves by issuing frequent "make check" commands), but it was certainly not considered a major failure.

In four words: git saved the day.
  Sam
-- 
Samuel Tardieu -- sam@rfc1149.net -- http://www.rfc1149.net/
Previous: Avery PennarunNext: Raymond Auge
Message 3 of 4 in “sadly requesting help”
  1. Raymond AugeMar 30, 2010
  2. Avery PennarunMar 30, 2010
  3. Samuel TardieuMar 31, 2010
  4. Fwd: sadly requesting helpRaymond Auge, Apr 2, 2010

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.