[sheepdog] [PATCH 1/2] dog: implement vdi check for erasure coded vdi
MORITA Kazutaka
morita.kazutaka at gmail.com
Fri Oct 18 20:23:45 CEST 2013
At Sat, 19 Oct 2013 00:26:37 +0800,
Liu Yuan wrote:
>
> When data copy is missing, this command will rebuild it.
>
> For the case that data get inconsistent but still accessable, we just print
> inconsistent warning because it is kind of hard to handle it and leave it
> as a future work
>
> Signed-off-by: Liu Yuan <namei.unix at gmail.com>
> ---
> dog/common.c | 9 +++
> dog/dog.h | 1 +
> dog/vdi.c | 148 ++++++++++++++++++++++++++++++++++++++--------
> tests/functional/050 | 2 +-
> tests/functional/052 | 22 +++----
> tests/functional/052.out | 128 ++++++++++++++++++++++++++++-----------
> tests/functional/055 | 8 +--
> tests/functional/055.out | 12 ++--
> tests/functional/056 | 12 ++--
> tests/functional/056.out | 10 ++--
> tests/functional/058 | 8 +--
> tests/functional/058.out | 3 +
> 12 files changed, 265 insertions(+), 98 deletions(-)
>
> diff --git a/dog/common.c b/dog/common.c
> index 028d367..5e7ce2e 100644
> --- a/dog/common.c
> +++ b/dog/common.c
> @@ -342,3 +342,12 @@ size_t get_store_objsize(uint8_t copy_policy, uint64_t oid)
> }
> return get_objsize(oid);
> }
> +
> +bool is_erasure_oid(uint64_t oid, uint8_t policy)
> +{
> + if (is_vdi_obj(oid))
> + return false;
> + if (policy == 0)
> + return false;
> + return true;
> +}
> diff --git a/dog/dog.h b/dog/dog.h
> index cfc9d54..769fc6c 100644
> --- a/dog/dog.h
> +++ b/dog/dog.h
> @@ -83,6 +83,7 @@ int do_vdi_create(const char *vdiname, int64_t vdi_size,
> int do_vdi_check(const struct sd_inode *inode);
> void show_progress(uint64_t done, uint64_t total, bool raw);
> size_t get_store_objsize(uint8_t copy_policy, uint64_t oid);
> +bool is_erasure_oid(uint64_t oid, uint8_t policy);
>
> extern struct command vdi_command;
> extern struct command node_command;
> diff --git a/dog/vdi.c b/dog/vdi.c
> index 2639007..4362f13 100644
> --- a/dog/vdi.c
> +++ b/dog/vdi.c
> @@ -1397,7 +1397,7 @@ static void *read_object_from(const struct sd_vnode *vnode, uint64_t oid)
> }
>
> static void write_object_to(const struct sd_vnode *vnode, uint64_t oid,
> - void *buf, bool create)
> + void *buf, bool create, uint8_t ec_index)
> {
> struct sd_req hdr;
> struct sd_rsp *rsp = (struct sd_rsp *)&hdr;
> @@ -1411,6 +1411,7 @@ static void write_object_to(const struct sd_vnode *vnode, uint64_t oid,
> hdr.flags = SD_FLAG_CMD_WRITE;
> hdr.data_length = get_objsize(oid);
> hdr.obj.oid = oid;
> + hdr.obj.ec_index = ec_index;
>
> ret = dog_exec_req(&vnode->node->nid, &hdr, buf);
> if (ret < 0)
> @@ -1427,13 +1428,16 @@ struct vdi_check_work {
> struct vdi_check_info *info;
> const struct sd_vnode *vnode;
> uint8_t hash[SHA1_DIGEST_SIZE];
> + uint8_t ec_index;
> + uint8_t *buf;
> bool object_found;
> struct work work;
> };
>
> struct vdi_check_info {
> uint64_t oid;
> - int nr_copies;
> + uint8_t nr_copies;
> + uint8_t copy_policy;
> uint64_t total;
> uint64_t *done;
> int refcnt;
> @@ -1459,7 +1463,7 @@ static void vdi_repair_work(struct work *work)
> void *buf;
>
> buf = read_object_from(info->base->vnode, info->oid);
> - write_object_to(vcw->vnode, info->oid, buf, !vcw->object_found);
> + write_object_to(vcw->vnode, info->oid, buf, !vcw->object_found, 0);
> free(buf);
> }
>
> @@ -1479,7 +1483,7 @@ static void vdi_repair_main(struct work *work)
> free_vdi_check_info(info);
> }
>
> -static void vdi_hash_check_work(struct work *work)
> +static void vdi_check_object_work(struct work *work)
> {
> struct vdi_check_work *vcw = container_of(work, struct vdi_check_work,
> work);
> @@ -1488,19 +1492,29 @@ static void vdi_hash_check_work(struct work *work)
> struct sd_req hdr;
> struct sd_rsp *rsp = (struct sd_rsp *)&hdr;
>
> - sd_init_req(&hdr, SD_OP_GET_HASH);
> + if (is_erasure_oid(info->oid, info->copy_policy)) {
> + sd_init_req(&hdr, SD_OP_READ_PEER);
> + hdr.data_length = get_store_objsize(info->copy_policy,
> + info->oid);
> + hdr.obj.ec_index = vcw->ec_index;
> + hdr.epoch = sd_epoch;
> + vcw->buf = xmalloc(hdr.data_length);
> + } else
> + sd_init_req(&hdr, SD_OP_GET_HASH);
> hdr.obj.oid = info->oid;
> hdr.obj.tgt_epoch = sd_epoch;
>
> - ret = dog_exec_req(&vcw->vnode->node->nid, &hdr, NULL);
> + ret = dog_exec_req(&vcw->vnode->node->nid, &hdr, vcw->buf);
> if (ret < 0)
> exit(EXIT_SYSFAIL);
>
> switch (rsp->result) {
> case SD_RES_SUCCESS:
> vcw->object_found = true;
> - memcpy(vcw->hash, rsp->hash.digest, sizeof(vcw->hash));
> - uatomic_set(&info->base, vcw);
> + if (!is_erasure_oid(info->oid, info->copy_policy)) {
> + memcpy(vcw->hash, rsp->hash.digest, sizeof(vcw->hash));
> + uatomic_set(&info->base, vcw);
> + }
> break;
> case SD_RES_NO_OBJ:
> vcw->object_found = false;
> @@ -1514,17 +1528,9 @@ static void vdi_hash_check_work(struct work *work)
> }
> }
>
> -static void vdi_hash_check_main(struct work *work)
> +static void check_replicatoin_object(struct vdi_check_info *info)
> {
> - struct vdi_check_work *vcw = container_of(work, struct vdi_check_work,
> - work);
> - struct vdi_check_info *info = vcw->info;
> -
> - info->refcnt--;
> - if (info->refcnt > 0)
> - return;
> -
> - if (info->base == NULL) {
> + if (info->base == NULL) {
> sd_err("no node has %" PRIx64, info->oid);
> exit(EXIT_FAILURE);
> }
> @@ -1542,6 +1548,99 @@ static void vdi_hash_check_main(struct work *work)
> queue_work(info->wq, &info->vcw[i].work);
> }
> }
> +}
> +
> +static void check_erasure_object(struct vdi_check_info *info)
> +{
> + int d = 0, p = 0, i, j, k;
> + int dp = ec_policy_to_dp(info->copy_policy, &d, &p);
> + struct fec *ctx = ec_init(d, dp);
> + int miss_idx[dp], input_idx[dp];
> + size_t strip_size = SD_EC_DATA_STRIPE_SIZE / d;
> + uint64_t oid = info->oid;
> + size_t len = get_store_objsize(info->copy_policy, oid);
> + char *obj = xmalloc(len);
> + uint8_t *input[dp];
> +
> + for (i = 0; i < dp; i++)
> + miss_idx[i] = -1;
> +
> + for (i = 0, j = 0, k = 0; i < info->nr_copies; i++)
> + if (!info->vcw[i].object_found) {
> + miss_idx[j++] = i;
> + } else {
> + input_idx[k] = i;
> + input[k] = info->vcw[i].buf;
> + k++;
> + }
> +
> + if (!j) { /* No object missing */
> + int idx[d];
> +
> + for (i = 0; i < d; i++)
> + idx[i] = i;
> +
> + for (k = 0; k < p; k++) {
> + for (i = 0; i < SD_EC_NR_STRIPE_PER_OBJECT; i++) {
> + const uint8_t *ds[d];
> + uint8_t out[strip_size];
> +
> + for (j = 0; j < d; j++)
> + ds[j] = info->vcw[j].buf + strip_size
> + * i;
> + ec_decode(ctx, ds, idx, out, d + k);
> + memcpy(obj + strip_size * i, out, strip_size);
> + }
> + if (memcmp(obj, info->vcw[d + k].buf, len) != 0) {
> + /* TODO repair the inconsistency */
> + sd_err("object %"PRIx64" is inconsistent", oid);
> + goto out;
> + }
> + }
> + } else if (j > p) {
> + sd_err("failed to rebuild object %"PRIx64". %d copies get "
> + "lost, more than %d", oid, j, p);
> + goto out;
> + } else {
> + for (k = 0; k < j; k++) {
> + int m = miss_idx[k], n;
> +
> + for (i = 0; i < SD_EC_NR_STRIPE_PER_OBJECT; i++) {
> + const uint8_t *ds[d];
> + uint8_t out[strip_size];
> +
> + for (n = 0; n < d; n++)
> + ds[n] = input[n] + strip_size * i;
> + ec_decode(ctx, ds, input_idx, out, m);
> + memcpy(obj + strip_size * i, out, strip_size);
> + }
> + write_object_to(info->vcw[m].vnode, oid, obj, true,
> + info->vcw[m].ec_index);
> + fprintf(stdout, "fixed missing %"PRIx64", "
> + "copy index %d\n", info->oid, m);
> + }
> + }
> +out:
> + for (i = 0; i < dp; i++)
> + free(info->vcw[i].buf);
> + free(obj);
> + ec_destroy(ctx);
> +}
> +
> +static void vdi_check_object_main(struct work *work)
> +{
> + struct vdi_check_work *vcw = container_of(work, struct vdi_check_work,
> + work);
> + struct vdi_check_info *info = vcw->info;
> +
> + info->refcnt--;
> + if (info->refcnt > 0)
> + return;
> +
> + if (is_erasure_oid(info->oid, info->copy_policy))
> + check_erasure_object(info);
> + else
> + check_replicatoin_object(info);
>
> if (info->refcnt == 0)
> free_vdi_check_info(info);
> @@ -1560,13 +1659,15 @@ static void queue_vdi_check_work(const struct sd_inode *inode, uint64_t oid,
> info->total = inode->vdi_size;
> info->done = done;
> info->wq = wq;
> + info->copy_policy = inode->copy_policy;
>
> oid_to_vnodes(oid, &sd_vroot, nr_copies, tgt_vnodes);
> for (int i = 0; i < nr_copies; i++) {
> info->vcw[i].info = info;
> + info->vcw[i].ec_index = i;
> info->vcw[i].vnode = tgt_vnodes[i];
> - info->vcw[i].work.fn = vdi_hash_check_work;
> - info->vcw[i].work.done = vdi_hash_check_main;
> + info->vcw[i].work.fn = vdi_check_object_work;
> + info->vcw[i].work.done = vdi_check_object_main;
> info->refcnt++;
> queue_work(info->wq, &info->vcw[i].work);
> }
> @@ -1586,6 +1687,8 @@ int do_vdi_check(const struct sd_inode *inode)
>
> wq = create_work_queue("vdi check", WQ_DYNAMIC);
>
> + init_fec();
> +
> queue_vdi_check_work(inode, vid_to_vdi_oid(inode->vdi_id), NULL, wq);
>
> max_idx = count_data_objs(inode);
> @@ -1622,11 +1725,6 @@ static int vdi_check(int argc, char **argv)
> return ret;
> }
>
> - if (inode->copy_policy > 0) {
> - sd_err("not implemented for erasure coded vdi");
> - return EXIT_FAILURE;
> - }
> -
> return do_vdi_check(inode);
> }
>
> diff --git a/tests/functional/050 b/tests/functional/050
> index ab28f1f..c7cfbab 100755
> --- a/tests/functional/050
> +++ b/tests/functional/050
> @@ -22,7 +22,7 @@ _wait_for_sheep 6
>
> _cluster_format -c 6
>
> -$DOG vdi create test 100M
> +_vdi_create test 100M
> dd if=/dev/zero | $DOG vdi write test &
>
> # simulate IO NIC down of sheep 1
> diff --git a/tests/functional/052 b/tests/functional/052
> index bff9d52..2d556eb 100755
> --- a/tests/functional/052
> +++ b/tests/functional/052
> @@ -4,22 +4,22 @@
>
> . ./common
>
> -for i in 0 1 2 3; do
> +for i in `seq 0 5`; do
> _start_sheep $i
> done
> -_wait_for_sheep 4
> -_cluster_format
> -$DOG vdi create test 20M -P
> +_wait_for_sheep 6
> +_cluster_format -c 6
> +_vdi_create test 20M -P
Rather than adding '-c 6' to _cluster_format, it looks cleaner to
update _vdi_create() so that it creates 6 copies when -ec is not
specified.
Thanks,
Kazutaka
More information about the sheepdog
mailing list