From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001 From: Wolfgang Bumiller Date: Tue, 10 Mar 2020 12:55:04 +0100 Subject: [PATCH 07/32] PVE: [Config] rbd: block: rbd: disable rbd_cache_writethrough_until_flush Either the cache mode asks for a cache or not. There's no point in having a "temporary" cache mode. This option AFAIK was introduced as a hack for ancient virtio drivers. If anything, we should have a separate option for it. Better yet, VMs affected by the related issue should simply explicitly choose writethrough. Signed-off-by: Wolfgang Bumiller Signed-off-by: Thomas Lamprecht --- block/rbd.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/block/rbd.c b/block/rbd.c index 027cbcc695..3ac7ff7bd5 100644 --- a/block/rbd.c +++ b/block/rbd.c @@ -637,6 +637,8 @@ static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx, rados_conf_set(*cluster, "rbd_cache", "false"); } + rados_conf_set(*cluster, "rbd_cache_writethrough_until_flush", "false"); + r = rados_connect(*cluster); if (r < 0) { error_setg_errno(errp, -r, "error connecting");