diff options
| author | Jeff Darcy <jdarcy@redhat.com> | 2016-12-08 16:24:15 -0500 |
|---|---|---|
| committer | Vijay Bellur <vbellur@redhat.com> | 2017-01-30 19:13:58 -0500 |
| commit | 1a95fc3036db51b82b6a80952f0908bc2019d24a (patch) | |
| tree | b983ac196a8165d5cb5e860a5ef97d3e9a41b5c9 /libglusterfs/src/event.h | |
| parent | 7f7d7a939e46b330a084d974451eee4757ba61b4 (diff) | |
core: run many bricks within one glusterfsd process
This patch adds support for multiple brick translator stacks running
in a single brick server process. This reduces our per-brick memory usage by
approximately 3x, and our appetite for TCP ports even more. It also creates
potential to avoid process/thread thrashing, and to improve QoS by scheduling
more carefully across the bricks, but realizing that potential will require
further work.
Multiplexing is controlled by the "cluster.brick-multiplex" global option. By
default it's off, and bricks are started in separate processes as before. If
multiplexing is enabled, then *compatible* bricks (mostly those with the same
transport options) will be started in the same process.
Change-Id: I45059454e51d6f4cbb29a4953359c09a408695cb
BUG: 1385758
Signed-off-by: Jeff Darcy <jdarcy@redhat.com>
Reviewed-on: https://review.gluster.org/14763
Smoke: Gluster Build System <jenkins@build.gluster.org>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
Diffstat (limited to 'libglusterfs/src/event.h')
| -rw-r--r-- | libglusterfs/src/event.h | 16 |
1 files changed, 15 insertions, 1 deletions
diff --git a/libglusterfs/src/event.h b/libglusterfs/src/event.h index b01ef24bb8e..1348f5d05c0 100644 --- a/libglusterfs/src/event.h +++ b/libglusterfs/src/event.h @@ -28,7 +28,7 @@ typedef int (*event_handler_t) (int fd, int idx, void *data, #define EVENT_EPOLL_TABLES 1024 #define EVENT_EPOLL_SLOTS 1024 -#define EVENT_MAX_THREADS 32 +#define EVENT_MAX_THREADS 1024 struct event_pool { struct event_ops *ops; @@ -57,6 +57,20 @@ struct event_pool { * and live status */ int destroy; int activethreadcount; + + /* + * Number of threads created by auto-scaling, *in addition to* the + * configured number of threads. This is only applicable on the + * server, where we try to keep the number of threads around the number + * of bricks. In that case, the configured number is just "extra" + * threads to handle requests in excess of one per brick (including + * requests on the GlusterD connection). For clients or GlusterD, this + * number will always be zero, so the "extra" is all we have. + * + * TBD: consider auto-scaling for clients as well + */ + int auto_thread_count; + }; struct event_ops { |