1. 04 Apr, 2011 2 commits
    • Leigh B Stoller's avatar
      Convert from Mysql to DBI; all of these scripts are non-critical and · b0b17ec9
      Leigh B Stoller authored
      they can just convert directly over.
    • Mike Hibler's avatar
      Bug fix: ensure we run any spawned frisbeed as "the appropriate user." · 7f775c7d
      Mike Hibler authored
      Previously, frisbeed's were always running as the uid/gid of the master server
      which generally meant root. This was not a security issue, as we did not rely
      on file permissions to enforce accessibility.
      For the Emulab config, "appropriate user" means the uid of the experiment
      swapper and the gid of the experiment. For the null (default) config, it
      still means the uid/gid of the master server--if you don't want them running
      as root, don't run the master server as root.
      For now, and frisbee clients spawned (e.g., to fetch an image from a parent)
      are still run with the uid/gid of the master server. This can easily be
      changed if necessary.
  2. 01 Apr, 2011 8 commits
  3. 31 Mar, 2011 17 commits
  4. 30 Mar, 2011 5 commits
  5. 29 Mar, 2011 4 commits
  6. 28 Mar, 2011 3 commits
    • Leigh B Stoller's avatar
    • Leigh B Stoller's avatar
    • Leigh B Stoller's avatar
      Begin the transition away from the ancient Mysql.pm module to the more · 5030b44d
      Leigh B Stoller authored
      current and maintained DBI::mysql module. A couple of things make this
      a little more work then you might think.
      Mysql exports a slightly different API then DBI, both at the DB *and*
      the statement level. The former required some restructuring of
      emdbi.pm, partly cause we want external sites to continue using Mysql
      for a while longer. So, emdbi suppports both interfaces, via the
      configure variable TBUSEDBI.
      I also took the opportunity to also scrap the existing fork()
      detection code and redo it in an easier to understand manner.
      Actually, I had no idea what the previous code was trying to do, so it
      was easier to just get rid of it, rather then try to make it work for
      the DBI API.
      There are also API differences in the "statement" class, but
      fortunately this can be hidden by wrapping the statement class with a
      wrapper that adds the routines we need to avoid making silly changes
      to 1000s of queries. They are all trivial little things since mostly
      its a matter of naming (numrows --> rows).
      I also changed the library we use on ops (db/libtbdb.pm.in) to use
      DBI, but in this case I just switched it over. Seemed like overkill to
      worry about supporting both APIs on ops. If it works it works, and so
      far it does. 
      Lastly, the following modules still use Mysql directly. They all need
      to be changed, but none of these are on the critical path to swapin
      and swapout, so they can change later.
  7. 26 Mar, 2011 1 commit