openstack/swift
Code Issues Proposed changes

Merge "Extend concurrent_gets to EC GET requests"

Browse Source
This commit is contained in:
Zuul
2020-09-04 14:22:19 +00:00
committed by Gerrit Code Review
parent a7f3fe93ab 8f60e0a260
commit 20e1544ad8
7 changed files with 875 additions and 336 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
etc/proxy-server.conf-sample
View File

@@ -200,12 +200,14 @@ use = egg:swift#proxy
# the number of seconds configured by timing_expiry.
# timing_expiry = 300
#
# By default on a GET/HEAD swift will connect to a storage node one at a time
# in a single thread. There is smarts in the order they are hit however. If you
# turn on concurrent_gets below, then replica count threads will be used.
# With addition of the concurrency_timeout option this will allow swift to send
# out GET/HEAD requests to the storage nodes concurrently and answer with the
# first to respond. With an EC policy the parameter only affects HEAD requests.
# By default on a GET/HEAD swift will connect to a minimum number storage nodes
# in a minimum number of threads - for replicated data just a single request to
# a single node one at a time. When enabled concurrent_gets allows the proxy,
# to use up to replica count threads when waiting on a response. In
# conjunction with the concurrency_timeout option this will allow swift to send
# out GET/HEAD requests to the storage nodes concurrently and answer as soon as
# the minimum number of backend responses are availabe - in replicated contexts
# this will be the first backend replica to respond.
# concurrent_gets = off
#
# This parameter controls how long to wait before firing off the next
@@ -215,6 +217,13 @@ use = egg:swift#proxy
# conn_timeout parameter.
# concurrency_timeout = 0.5
#
# By default on a EC GET request swift will connect to a minimum number of
# storage nodes in a minimum number of threads - for erasure coded data, ndata
# requests to primary nodes are started at the same time. When greater than
# zero this option provides additional robustness and may reduce first byte
# latency by starting additional requests - up to as many as nparity.
# concurrent_ec_extra_requests = 0
#
# Set to the number of nodes to contact for a normal request. You can use
# '* replicas' at the end to have it use the number given times the number of
# replicas for the ring being used for the request.