[ARVADOS] updated: 2.1.0-2178-g0a9d4fa50
Git user
git at public.arvados.org
Wed Mar 30 17:28:45 UTC 2022
Summary of changes:
services/fuse/arvados_fuse/fusedir.py | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
via 0a9d4fa5043bd7291611c41588bdd3f0b70ede44 (commit)
from d3051d45df5ee760a05a84bf4d385799c4326477 (commit)
Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.
commit 0a9d4fa5043bd7291611c41588bdd3f0b70ede44
Author: Peter Amstutz <peter.amstutz at curii.com>
Date: Wed Mar 30 13:27:55 2022 -0400
18941: Tweak the prefetch thread max to 7
Arvados-DCO-1.1-Signed-off-by: Peter Amstutz <peter.amstutz at curii.com>
diff --git a/services/fuse/arvados_fuse/fusedir.py b/services/fuse/arvados_fuse/fusedir.py
index 876825a37..f3816c0d3 100644
--- a/services/fuse/arvados_fuse/fusedir.py
+++ b/services/fuse/arvados_fuse/fusedir.py
@@ -525,9 +525,12 @@ class CollectionDirectory(CollectionDirectoryBase):
self.collection.update()
new_collection_record = self.collection.api_response()
else:
- # too much prefetch and you end up stepping on your own transfers
- # experimentally the optimal somewhere between 4 and 6
- get_threads = min(max((self.api.keep.block_cache.cache_max // (64 * 1024 * 1024)) - 1, 1), 6)
+ # If there's too many prefetch threads and you
+ # max out the CPU, delivering data to the FUSE
+ # layer actually ends up being slower.
+ # Experimentally, capping 7 threads seems to
+ # be a sweet spot.
+ get_threads = min(max((self.api.keep.block_cache.cache_max // (64 * 1024 * 1024)) - 1, 1), 7)
# Create a new collection object
if uuid_pattern.match(self.collection_locator):
coll_reader = arvados.collection.Collection(
-----------------------------------------------------------------------
hooks/post-receive
--
More information about the arvados-commits
mailing list