openstack/oslo.service
Code Issues Proposed changes

wsgi: decrease the default number of greenthreads in pool

Browse Source 
It turned out that distribution of WSGI requests between forks is not
even when eventlet is used, which can cause pool timeout issues under
load, while there are idle workers with open DB connections, which
should have served these HTTP requests.

The default number of greenthreads in pool should be decreased to allow
DB oriented services (mostly APIs) distribute load more evenly between
workers: one worker will stop accepting new requests, when there are no
more available greenthreads in the wsgi pool, accept() will block, but
there are other forks accept()'ing on the same FD to handle connections
from the queue in kernel.

Related change to oslo.db:

  I2e9c2a71d8231e0dfbefc6293ad319e1e459beec

Testing was performed using this script:

  https://gist.github.com/zzzeek/c69138fd0d0b3e553a1f

With 100 greenthreads in the pool, 50 DB connections allowed overflow
no pool timeout issues were seen with up to 500 concurrent requests
done by ab, while current default values (1000/10) could not handle
even 100 concurrent requests.

See this ML thread for details:

  http://lists.openstack.org/pipermail/openstack-dev/2015-December/082717.html

Closes-Bug: #1535375

Change-Id: I65b40b9906b75146a0085bbe168f1e6bcae82f21
This commit is contained in:
Roman Podoliaka 2016-01-18 18:44:33 +02:00
parent 01d747aa95
commit 321c185f3d
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
oslo_service/_options.py
View File

@ -69,7 +69,7 @@ wsgi_opts = [
help="Sets the value of TCP_KEEPIDLE in seconds for each "
"server socket. Not supported on OS X."),
cfg.IntOpt('wsgi_default_pool_size',
default=1000,
default=100,
help="Size of the pool of greenthreads used by wsgi"),
cfg.IntOpt('max_header_line',
default=16384,