[tahoe-lafs-trac-stream] [tahoe-lafs] #1870: leasedb: performance regression

[tahoe-lafs]

#1870: leasedb: performance regression
-------------------------+-------------------------------------------------
     Reporter:           |      Owner:  daira
  davidsarah             |     Status:  new
         Type:  defect   |  Milestone:  1.11.0
     Priority:  normal   |    Version:  1.9.2
    Component:  code-    |   Keywords:  leasedb performance regression
  storage                |  sqlite
   Resolution:           |
Launchpad Bug:           |
-------------------------+-------------------------------------------------

Comment (by daira):

 Replying to [comment:33 warner]:
 > As we noticed in the dev-chat today, `Cp.test_copy_using_filecap` is
 mistakenly copying its entire basedir into the tahoe filesystem (which
 includes all 10 server basedirs, and their shares, which are being created
 during the `cp` by the files being copied in.. it's remarkable that the
 process terminates at all). This resulted in something like 174 shares
 being created (per server), instead of 4. Fixing this doesn't address the
 overall regression, but does explain why this one test was taking 17
 seconds even on trunk.

 Fixed in #2048.

 > OTOH, leaving that test broken while we figure out the performance
 regression provides us with a convenient lots-of-shares stress test,
 that's highly responsive to changes in the leasedb times. I'd much rather
 iterate on a 30-ish second test than on the full test suite.

 That's fine, we'll just not pull that fix onto 1819-cloud-merge yet.

-- 
Ticket URL: <https://tahoe-lafs.org/trac/tahoe-lafs/ticket/1870#comment:34>
tahoe-lafs <https://tahoe-lafs.org>
secure decentralized storage