From 8c2ed94e9cbc7b1ed84b034a65b126e16b18a596 Mon Sep 17 00:00:00 2001 From: Ronen Friedman Date: Fri, 6 Sep 2024 15:49:12 +0300 Subject: [PATCH] osd/scrub: decrease default deep scrub chunk size The previous default of 25 objects per chunk proved to take too long (many hundreds of milliseconds) on a busy cluster. As the scrubber locks all objects in the chunk for the duration, a large chunk size can cause a significant impact on the client ops' latencies. Fixes: https://tracker.ceph.com/issues/68057 Signed-off-by: Ronen Friedman --- src/common/options/osd.yaml.in | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/src/common/options/osd.yaml.in b/src/common/options/osd.yaml.in index 268a89154de5a..6cd0a8a890472 100644 --- a/src/common/options/osd.yaml.in +++ b/src/common/options/osd.yaml.in @@ -349,8 +349,10 @@ options: type: int level: advanced desc: Maximum number of objects to deep-scrub in a single chunk - fmt_desc: The maximum number of object store chunks to scrub during single operation. - default: 25 + fmt_desc: The maximum number of objects to deep-scrub during single internal + scrub operation. Large values would improve scrubbing performance but + may adversely affect client operations' latency. + default: 15 see_also: - osd_scrub_chunk_min with_legacy: true -- 2.39.5