Project

General

Profile

Support #49489

Getting Long heartbeat and slow requests on ceph luminous 12.2.13

Added by ceph ceph 9 months ago. Updated 7 months ago.

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

0%

Tags:
Long hearbeat, NVME cached SSD pool, osd flaps
Reviewed:
Affected Versions:
Component(RADOS):
Pull request ID:

Description

1. Current environment is integrated with ceph and openstack
2. It has NVME and SSD disks only
3. We have create following pool - 1. NVME cached ssd where NVME acts are cache tier and SSD as main storage
2. Pure SSD pool
3. Pure NVME pool

4. Creating vms from pure ssd and nvme has no issues
5. Creating vms from NVME cached SSD showing long heartbeats on back and front network as well as slow requests
6. Pool are in write back mode
7. Then we are also seeing OSD flaps, mon down also

WhatsApp Image 2021-02-22 at 9.15.48 PM.jpeg View - actual screenshot from ceph (49.3 KB) ceph ceph, 02/25/2021 09:23 AM

History

#1 Updated by Greg Farnum 7 months ago

  • Tracker changed from Bug to Support
  • Project changed from Ceph to RADOS
  • Category deleted (OSD)

This is almost certainly a result of cache tiering (which we generally discourage from use) being a bad fit or incorrectly configured for the workload.

Also available in: Atom PDF