[DRBD-user] drbd 8.4.7 out-of-sync on disabled host page cache in VM

Armin Schindler Armin.Schindler at sysgo.com
Mon Mar 11 11:13:11 CET 2019


Hello,

we often get out-of-sync blocks when using 'verify' and we didn't find a
solution yet. Searching old posts didn't help so far.

The setup is:
2 hosts Debian 9 (stretch) with default DRBD version 8.4.7.
Both systems have hardware RAID5 and one big partition is the physical volume
for LVM.
A DRBD volume is configured with a LV as disk and another LV (size 128M) as
meta-disk.

DRBD devices are used for local mount for LXC type containers root-fs or
as virtual disk for KVM/Qemu VMs.

drbd verify runs weekly and often shows out-of-sync blocks. A
disconnect->connect seems to fix that status then.

After months with testing different settings, the out-of-sync status does
not occur when KVM/Qemu is used with a cache setting of enabled host page
cache (writethrough, writeback, unsafe).

When using KVM/Qemu with cache mode of disabled host page cache (none,
directsync) out-of-sync happens.

Also, when using DRBD volume without KVM/Qemu and mounted locally e.g.
as ext4 filesystem (and a started LXC container on it), out of sync
happens too.

In other threads the out-of-sync situation is often related to KVM/Qemu, but
this is not true. I can reproduce it on the host itself.

Thank you for any help or ideas.
Armin




-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2900 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20190311/ce42e45d/attachment.bin>


More information about the drbd-user mailing list