Tim Burke 452db14a09 Bind a new socket per-worker ...

We've seen lumpy distributions of requests to workers, which seems to
parallel what some other projects have seen [0][1].

The solution (as best we can tell) is to take advantage of the
SO_REUSEPORT that eventlet's been setting for us since basically
forever [2].

[0] https://lwn.net/Articles/542629/
[1] https://github.com/varnish/hitch/issues/142
[2] https://github.com/eventlet/eventlet/commit/f9a3074a3

Change-Id: I83cdaa2cbd394cbd49609c65bf9c5ed026c55417

2020-08-20 21:03:37 -07:00

..

account

Allow direct and internal clients to use the replication network

2020-08-04 21:22:04 +00:00

cli

Merge "Treat add/remove/rebalance during part-power increase as error"

2020-07-25 06:48:05 +00:00

common

Bind a new socket per-worker

2020-08-20 21:03:37 -07:00

container

Merge "Address a sharder/replicator race"

2020-08-07 07:46:19 +00:00

obj

Allow replication servers to handle all request methods

2020-07-23 09:11:07 -07:00

proxy

Allow direct and internal clients to use the replication network

2020-08-04 21:22:04 +00:00

test_locale

py3: port the test of locale

2019-06-19 09:54:14 -07:00

__init__.py

Allow direct and internal clients to use the replication network

2020-08-04 21:22:04 +00:00

helpers.py

Allow internal clients to use reserved namespace

2019-11-27 11:22:00 -06:00