|
From: | Vladimir Sementsov-Ogievskiy |
Subject: | [Qemu-block] write_zeroes/trim on the whole disk |
Date: | Fri, 23 Sep 2016 21:32:12 +0300 |
User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.6.0 |
Hi all! There is a following problem. When we need to write_zeroes or trim the whole disk, we have to do it iteratively, because of 32-bit restriction on request length. For example, current implementation of mirror (see mirror_dirty_init()) do this by chunks of 2147418112 bytes (with default granularity of 65536). So, to zero 16tb disk we will make 8192 requests instead of one. Incremental zeroing of 1tb qcow2 takes > 80 seconds for me (see below). This means ~20 minutes for copying empty 16tb qcow2 disk which is obviously a waste of time. We see the following solutions for nbd:
1. Add command NBD_MAKE_EMPTY, with flag, saying what should be done: trim or write_zeroes. 2. Add flag NBD_CMD_FLAG_WHOLE for commands NBD_TRIM and NBD_WRITE_ZEROES, which will say (with zeroed offset and lenght of the request), that the whole disk should be discarded/zeroed. 3. Increase length field of the request to 64bit. As soon as we have some way to empty disk in nbd, we can use qcow2_make_empty, to trim the whole disk (and something similar should be done for zeroing). What do you think about this all, and which way has a chance to get into nbd proto? == test incremental qcow2 zeroing in mirror == 1. enable it. If we will use nbd, it will be enabled. diff --git a/block/mirror.c b/block/mirror.c index f9d1fec..4ac0c39 100644 --- a/block/mirror.c +++ b/block/mirror.c @@ -556,7 +556,7 @@ static int coroutine_fn mirror_dirty_init(MirrorBlockJob *s) end = s->bdev_length / BDRV_SECTOR_SIZE; - if (base == NULL && !bdrv_has_zero_init(target_bs)) { + if (base == NULL) { if (!bdrv_can_write_zeroes_with_unmap(target_bs)) { bdrv_set_dirty_bitmap(s->dirty_bitmap, 0, end); return 0; ==== test ==== qemu-img create -f qcow2 /tmp/1tb.qcow2 1T virsh start backup-vm --paused Domain backup-vm started virsh qemu-monitor-command backup-vm {"execute":"blockdev-add","arguments":{"options": {"aio": "native", "file": {"driver": "file", "filename": "/tmp/1tb.qcow2"}, "discard": "unmap", "cache": {"direct": true}, "driver": "qcow2", "id": "disk"}}} {"return":{},"id":"libvirt-32"} /usr/bin/time -f '%e seconds' sh -c 'virsh qemu-monitor-event' & virsh qemu-monitor-command backup-vm {"execute":"drive-mirror","arguments":{"device": "disk", "sync": "full", "target": "/tmp/targ"}} {"return":{},"id":"libvirt-33"} address@hidden qemu]# event BLOCK_JOB_READY at 1474652677.668624 for domain backup-vm: {"device":"disk","len":1099511627776,"offset":1099511627776,"speed":0,"type":"mirror"} events received: 1 86.39 seconds - the same for 2tb empty disk: 180.19 seconds - and without patch, it takes < 1 second, of course. -- Best regards, Vladimir |
[Prev in Thread] | Current Thread | [Next in Thread] |