Project

General

Profile

Actions

Bug #23008

closed

Filestore rocksdb compaction readahead option not set by default

Added by Nathan Cutler about 6 years ago. Updated about 6 years ago.

Status:
Duplicate
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
% Done:

0%

Source:
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

See: https://github.com/facebook/rocksdb/blob/master/include/rocksdb/options.h#L680

While RocksDB code comments mention HDDs specifically, we also saw improved CPU usage due to fewer read calls with readahead enabled on NVMe (Atleast with BlueFS, milleage may vary with filestore and rocksdb using posix). In any event, reads are all entirely sequential as verified by blktrace so there doesn't appear to be any real downside to enabling this.


Related issues 1 (0 open1 closed)

Is duplicate of Ceph - Bug #21505: filestore rocksdb compaction readaheadResolved09/22/2017

Actions
Actions #2

Updated by Nathan Cutler about 6 years ago

  • Copied to Backport #23009: luminous: Filestore rocksdb compaction readahead option not set by default added
Actions #3

Updated by Nathan Cutler about 6 years ago

  • Copied to Backport #23010: jewel: Filestore rocksdb compaction readahead option not set by default added
Actions #4

Updated by Nathan Cutler about 6 years ago

  • Copied to deleted (Backport #23009: luminous: Filestore rocksdb compaction readahead option not set by default)
Actions #5

Updated by Nathan Cutler about 6 years ago

  • Copied to deleted (Backport #23010: jewel: Filestore rocksdb compaction readahead option not set by default)
Actions #6

Updated by Nathan Cutler about 6 years ago

  • Status changed from Pending Backport to Duplicate
  • Backport deleted (luminous, jewel)
Actions #7

Updated by Nathan Cutler about 6 years ago

  • Is duplicate of Bug #21505: filestore rocksdb compaction readahead added
Actions

Also available in: Atom PDF