search

lipinggm / tlb

Automatically exported from code.google.com/p/tlb
0 stars 0 forks source link

Need to clear cert stores created in temp directory #60

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago 
For every test run, tlb balancer creates a trust store in the temp directory 
and over a period of many runs there are multiple folders in the /tmp directory 
(although they are not big files)

We need to clear these on a regular basis.

Original issue reported on code.google.com by mrajesh...@gmail.com on 1 Mar 2011 at 10:58

GoogleCodeExporter commented 9 years ago 
Make sure we create these file with deleteOnExit so that they get deleted when 
the VM exits. Since they are small, we will not add a periodic cleaner on the 
balancer side. Not until it becomes an issue.

Original comment by itspa...@gmail.com on 17 Mar 2011 at 8:20

GoogleCodeExporter commented 9 years ago

Original comment by itspa...@gmail.com on 7 Oct 2011 at 8:33

GoogleCodeExporter commented 9 years ago

Original comment by itspa...@gmail.com on 2 Nov 2011 at 4:48

GoogleCodeExporter commented 9 years ago 
Could not reproduce the cert store. However, came across a whole bunch of temp 
files that can be cleared out. Deleting them instead.

Original comment by itspa...@gmail.com on 2 Nov 2011 at 4:49

GoogleCodeExporter commented 9 years ago 
Closing this as fixed for now. If this comes up as an issue again, will revisit 
it.

Original comment by itspa...@gmail.com on 2 Nov 2011 at 5:42

GoogleCodeExporter commented 9 years ago 
Reopening this because forceDeleteOnExit causes memory leak. 

As a part of this we should figure out locking using lock-file to maintain a 
record of all files that we create and then knock off the files created by 
previous run(process id can be used to find out previous runs).

Or

We can create a sub-directory which is tlb managed, which can be deleted before 
exit. This directory name however, should not be hardcoded. It should be a UUID 
or a digest generated from environment variable configuration, so multiple TLB 
enabled builds can run on a single machine at a time.

Original comment by singh.janmejay on 8 Nov 2011 at 7:14

GoogleCodeExporter commented 9 years ago

Original comment by itspa...@gmail.com on 20 Nov 2011 at 7:25

GoogleCodeExporter commented 9 years ago 
Investigating. Not sure if this is still a problem.

Original comment by singh.janmejay on 24 Dec 2011 at 2:27

GoogleCodeExporter commented 9 years ago 
It was still a problem. Fix is to batch server posts in balancer(like we do 
when working against go-server), and clearing cache files when posting to the 
server eventually.

Fix is upstream, for some reason doesn't work too well with test-unit example. 
Need to investigate it before closing this one.

Original comment by singh.janmejay on 3 Jan 2012 at 1:44

GoogleCodeExporter commented 9 years ago 
Fixed test::unit support, the problem was rake exposing an entity of its own 
called rake_test_loader which runs like a test(has test startup and stop 
lifecycle events that test-listener listens to). This was causing an offset of 
one between what tlb expected as number of tests run and actual test runs that 
listener got callbacks for. This is stupid. 

The fix is to maintain the list of tests that balancer decides to run on the 
partition and only report balancer about test-times and results of those.

fixed in: 4ef3e7f5a386314519751cd2f3b68ec9fc55d3a3 (tlb.rb)

Original comment by singh.janmejay on 4 Jan 2012 at 4:48

GoogleCodeExporter commented 9 years ago 
TLB's own set of tests(especially tlb.rb tests) still leave stray /tmp files, 
but this hurts only tlb developers, so not a problem.

Original comment by singh.janmejay on 4 Jan 2012 at 4:49