At Mon, 28 Jul 2014 14:55:20 +0800,
Ruoyu wrote:
>
> If cluster members change frequently, especially the first two or
> more nodes of the cluster leave and join frequently, the following
> error message will be found in sheep.log
>
> Failed to find requested tag, remote address: 127.0.0.1:7001,
> op name: GET_EPOCH
>
> This patch adds epoch recovery to reduce these errors. To achieve
> it, it is better to unify epoch file because rb_node is unused
> in it. And then, we can create a new epoch file as long as the file
> is not found in local node.
>
> v2 is rebased on the latest master and
> 1. memset's 2nd parameter, use 0 instead of '\0'
> 2. for loop initial declaration
>
> Signed-off-by: Ruoyu <[email protected]>
> ---
> sheep/group.c | 7 +++++--
> sheep/store.c | 9 +++++++++
> 2 files changed, 14 insertions(+), 2 deletions(-)
Applied, thanks.
Hitoshi
>
> diff --git a/sheep/group.c b/sheep/group.c
> index 6b1ec5d..f53ad0f 100644
> --- a/sheep/group.c
> +++ b/sheep/group.c
> @@ -359,7 +359,7 @@ int epoch_log_read_remote(uint32_t epoch, struct sd_node
> *nodes, int len,
> rb_for_each_entry(node, &vinfo->nroot, rb) {
> struct sd_req hdr;
> struct sd_rsp *rsp = (struct sd_rsp *)&hdr;
> - int nodes_len;
> + int nodes_len, nr_nodes;
>
> if (node_is_local(node))
> continue;
> @@ -377,7 +377,10 @@ int epoch_log_read_remote(uint32_t epoch, struct sd_node
> *nodes, int len,
> if (timestamp)
> memcpy(timestamp, buf + nodes_len, sizeof(*timestamp));
>
> - return nodes_len / sizeof(struct sd_node);
> + nr_nodes = nodes_len / sizeof(struct sd_node);
> + /* epoch file is missing in local node, try to create one */
> + update_epoch_log(epoch, nodes, nr_nodes);
> + return nr_nodes;
> }
>
> /*
> diff --git a/sheep/store.c b/sheep/store.c
> index eb42f2c..ea445fc 100644
> --- a/sheep/store.c
> +++ b/sheep/store.c
> @@ -33,6 +33,15 @@ int update_epoch_log(uint32_t epoch, struct sd_node
> *nodes, size_t nr_nodes)
> memcpy(buf, nodes, nodes_len);
> memcpy(buf + nodes_len, &t, sizeof(time_t));
>
> + /*
> + * rb field is unused in epoch file, zero-filling it
> + * is good for epoch file recovery because it is unified
> + */
> + for (int i = 0; i < nr_nodes; i++)
> + memset(buf + i * sizeof(struct sd_node)
> + + offsetof(struct sd_node, rb),
> + 0, sizeof(struct rb_node));
> +
> snprintf(path, sizeof(path), "%s%08u", epoch_path, epoch);
>
> ret = atomic_create_and_write(path, buf, len, true);
> --
> 1.8.3.2
>
>
> --
> sheepdog mailing list
> [email protected]
> http://lists.wpkg.org/mailman/listinfo/sheepdog
--
sheepdog mailing list
[email protected]
http://lists.wpkg.org/mailman/listinfo/sheepdog
