tree 4b71d8d110cd272fad666aa01e369e3f5e3223fb
parent aed3196884f3c0529e16b7ce8741eacfe1244b0e
author Keith Turner <kturner@apache.org> 1549656306 -0500
committer GitHub <noreply@github.com> 1549656306 -0500
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJcXeDyCRBK7hj4Ov3rIwAAdHIIADc0mhJlySTU3Afst4duqszY
 klC4N+isrMhwtSMQ6udn7Ehz3IZ1SLiVKzW1W1XP1pX/Xbl1uFzxuRTiCF06HWNb
 K9aiXWt9D7IyoBJiT2g1/wkRRshM3Rw4ZqqmH2l3XdxlMf1Rvaw6+y7rfnkri0Ow
 PWGNuE0ke+R/8r5onZdJYRptD9WgOLJtyZUJ0MzIkVvPKqSLW5YAwGwsMF6OWAGZ
 tNhbihOcpo9o3mP1IeLcrsBDYNE00VJVAmnX17K03N/0FI/0nHkTUoRnq+mg6HIT
 dT7UltCW2GGa0GTpdgeSVnow+1NG1TC7q9pOOZj8dzghp5MnMvVU3iQThPvuqOs=
 =zqjl
 -----END PGP SIGNATURE-----
 

Added perf test for yielding+scan dispatching (#56)

Test running lots of short scans while many filters that return little
data are running in the background.  If nothing is done these filters
will prevent the short scans from running.  This test configures
Accumulo so that the short scans should get a chance to run.  It does
three things to facilitate the short scans : the filters yield, there
are two scan executors, and a scan dispatcher sends long running scans
to the second executor. If yielding and dispatching are working
correctly then the short scans should have very short response times.
This happens because the filters should end up in a separate thread pool
than the short scan.

Also fixed some bugs with the performance test framework.