Project

General

Profile

Actions

Bug #47873

closed

/usr/lib/sysctl.d/90-ceph-osd.conf getting installed in container, rendering it ineffective

Added by Nathan Cutler over 3 years ago. Updated almost 3 years ago.

Status:
Resolved
Priority:
High
Category:
cephadm (binary)
Target version:
-
% Done:

0%

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

Description

The file /usr/lib/sysctl.d/90-ceph-osd.conf has the following contents:

fs.aio-max-nr = 1048576
kernel.pid_max = 4194304

This is very important to have on all OSD nodes. However, since this file is owned by the ceph-osd package, and that package is installed in the container image, and not in the host system, it no longer takes effect. This causes problems on OSDs nodes with larger numbers of disks.

One option would be for cephadm to manage this file, ensuring that it gets created on OSD nodes when OSDs are created, and removing it when a node no longer has any OSDs.

Another option would be to move the file to ceph-base. Then it would be installed (in the host system) on all cluster nodes, regardless of whether the node has OSDs or not.

Yet another option would be to move the file into ceph-common. In this scenario, the file would be installed on both cluster nodes and client nodes as well.


Related issues 4 (0 open4 closed)

Related to Orchestrator - Feature #49171: cephadm: set osd-memory-targetResolvedSage Weil

Actions
Related to Orchestrator - Bug #50551: Massive OSD creation: kernel parameter fs.aio-max-nr with a low value by defaultDuplicateJuan Miguel Olmo Martínez

Actions
Related to Orchestrator - Feature #48292: cephadm: allow more than 60 OSDs per hostResolvedSebastian Wagner

Actions
Related to Orchestrator - Bug #50526: OSD massive creation: OSDs not createdResolvedJuan Miguel Olmo Martínez

Actions
Actions

Also available in: Atom PDF