diff options
author | Mohammed Junaid <junaid@redhat.com> | 2012-03-19 21:26:56 +0530 |
---|---|---|
committer | Vijay Bellur <vijay@gluster.com> | 2012-03-19 09:16:23 -0700 |
commit | 2ffefd720a54fb815b1efa11e9de766fe1518831 (patch) | |
tree | 1af16aff8fefe2184f7241e0f1c03433a3a49399 /xlators/protocol | |
parent | d98c3e19342be3b8b3e2c7e3d7e544de34143bdf (diff) |
protocol/client: Handle failures in lock self healing gracefully (part2).
During reopening of fd's and reacquiring of locks on the fd (after a
reconnect), a release on a fd on which reacquiring of locks is in progress
will free up fdctx. This patch will keep fdctx valid until the reacquiring
of locks is in progress.
Change-Id: I0fae27544a7f8ddaa26def4ee4d41a8a2b322521
BUG: 795386
Signed-off-by: Mohammed Junaid <junaid@redhat.com>
Reviewed-on: http://review.gluster.com/2819
Reviewed-by: Amar Tumballi <amarts@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Vijay Bellur <vijay@gluster.com>
Diffstat (limited to 'xlators/protocol')
-rw-r--r-- | xlators/protocol/client/src/client-handshake.c | 218 |
1 files changed, 169 insertions, 49 deletions
diff --git a/xlators/protocol/client/src/client-handshake.c b/xlators/protocol/client/src/client-handshake.c index 8b819a76f35..0a8759af586 100644 --- a/xlators/protocol/client/src/client-handshake.c +++ b/xlators/protocol/client/src/client-handshake.c @@ -406,6 +406,31 @@ client_notify_parents_child_up (xlator_t *this) } int +clnt_fd_lk_reacquire_failed (xlator_t *this, clnt_fd_ctx_t *fdctx, + clnt_conf_t *conf) +{ + int ret = -1; + + GF_VALIDATE_OR_GOTO ("client", this, out); + GF_VALIDATE_OR_GOTO (this->name, conf, out); + GF_VALIDATE_OR_GOTO (this->name, fdctx, out); + + pthread_mutex_lock (&conf->lock); + { + fdctx->remote_fd = -1; + fdctx->lk_heal_state = GF_LK_HEAL_DONE; + + list_add_tail (&fdctx->sfd_pos, + &conf->saved_fds); + } + pthread_mutex_unlock (&conf->lock); + + ret = 0; +out: + return ret; +} + +int client_set_lk_version_cbk (struct rpc_req *req, struct iovec *iov, int count, void *myframe) { @@ -427,11 +452,12 @@ client_set_lk_version_cbk (struct rpc_req *req, struct iovec *iov, gf_log (fr->this->name, GF_LOG_WARNING, "xdr decoding failed"); else - gf_log (fr->this->name, GF_LOG_DEBUG, + gf_log (fr->this->name, GF_LOG_INFO, "Server lk version = %d", rsp.lk_ver); ret = 0; out: + //TODO: Check for all released fdctx and destroy them if (fr) STACK_DESTROY (fr->root); @@ -464,9 +490,11 @@ client_set_lk_version (xlator_t *this) NULL, NULL, 0, NULL, 0, NULL, (xdrproc_t)xdr_gf_set_lk_ver_req); out: - if (ret < 0) + if (ret < 0) { + //TODO: Check for all released fdctx and destroy them gf_log (this->name, GF_LOG_WARNING, "Failed to send SET_LK_VERSION to server"); + } return ret; } @@ -522,7 +550,6 @@ clnt_fd_lk_local_unref (xlator_t *this, clnt_fd_lk_local_t *local) LOCK_DESTROY (&local->lock); GF_FREE (local); } - ref = 0; out: return ref; } @@ -556,7 +583,6 @@ clnt_mark_fd_bad (clnt_conf_t *conf, clnt_fd_ctx_t *fdctx) pthread_mutex_unlock (&conf->lock); } -// call decrement_reopen_fd_count int clnt_release_reopen_fd_cbk (struct rpc_req *req, struct iovec *iov, int count, void *myframe) @@ -571,7 +597,7 @@ clnt_release_reopen_fd_cbk (struct rpc_req *req, struct iovec *iov, fdctx = (clnt_fd_ctx_t *) frame->local; conf = (clnt_conf_t *) this->private; - clnt_mark_fd_bad (conf, fdctx); + clnt_fd_lk_reacquire_failed (this, fdctx, conf); decrement_reopen_fd_count (this, conf); @@ -603,16 +629,63 @@ clnt_release_reopen_fd (xlator_t *this, clnt_fd_ctx_t *fdctx) clnt_release_reopen_fd_cbk, NULL, NULL, 0, NULL, 0, NULL, (xdrproc_t)xdr_gfs3_releasedir_req); -out: + out: + if (ret) { + decrement_reopen_fd_count (this, conf); + clnt_fd_lk_reacquire_failed (this, fdctx, conf); + if (frame) { + frame->local = NULL; + STACK_DESTROY (frame->root); + } + } return 0; } int +clnt_reacquire_lock_error (xlator_t *this, clnt_fd_ctx_t *fdctx, + clnt_conf_t *conf) +{ + int32_t ret = -1; + + GF_VALIDATE_OR_GOTO ("client", this, out); + GF_VALIDATE_OR_GOTO (this->name, fdctx, out); + GF_VALIDATE_OR_GOTO (this->name, conf, out); + + clnt_release_reopen_fd (this, fdctx); + + ret = 0; +out: + return ret; +} + +gf_boolean_t +clnt_fd_lk_local_error_status (xlator_t *this, + clnt_fd_lk_local_t *local) +{ + gf_boolean_t error = _gf_false; + + LOCK (&local->lock); + { + error = local->error; + } + UNLOCK (&local->lock); + + return error; +} + +int clnt_fd_lk_local_mark_error (xlator_t *this, clnt_fd_lk_local_t *local) { + int32_t ret = -1; + clnt_conf_t *conf = NULL; gf_boolean_t error = _gf_false; + GF_VALIDATE_OR_GOTO ("client", this, out); + GF_VALIDATE_OR_GOTO (this->name, local, out); + + conf = (clnt_conf_t *) this->private; + LOCK (&local->lock); { error = local->error; @@ -620,30 +693,30 @@ clnt_fd_lk_local_mark_error (xlator_t *this, } UNLOCK (&local->lock); - if (error) - clnt_release_reopen_fd (this, local->fdctx); - - return 0; + if (!error) + clnt_reacquire_lock_error (this, local->fdctx, conf); + ret = 0; +out: + return ret; } -// Also, I think in reopen_cbk, the fdctx is added to -// saved_fd list.. avoid that, may cause a problem -// Reason: While the locks on the fd are reacquired, a release -// fop may be received by the client-protocol translator -// which will free the fdctx datastructure. int client_reacquire_lock_cbk (struct rpc_req *req, struct iovec *iov, int count, void *myframe) { int32_t ret = -1; xlator_t *this = NULL; - gf_common_rsp rsp = {0,}; + gfs3_lk_rsp rsp = {0,}; call_frame_t *frame = NULL; + clnt_conf_t *conf = NULL; + clnt_fd_ctx_t *fdctx = NULL; clnt_fd_lk_local_t *local = NULL; + struct gf_flock lock = {0,}; frame = (call_frame_t *) myframe; this = frame->this; local = (clnt_fd_lk_local_t *) frame->local; + conf = (clnt_conf_t *) this->private; if (req->rpc_status == -1) { gf_log ("client", GF_LOG_WARNING, @@ -651,7 +724,7 @@ client_reacquire_lock_cbk (struct rpc_req *req, struct iovec *iov, goto out; } - ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gf_common_rsp); + ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gfs3_lk_rsp); if (ret < 0) { gf_log (this->name, GF_LOG_ERROR, "XDR decoding failed"); goto out; @@ -663,15 +736,37 @@ client_reacquire_lock_cbk (struct rpc_req *req, struct iovec *iov, goto out; } - // TODO: Add more info to log. - gf_log (this->name, GF_LOG_DEBUG, "Reacquired lock"); + fdctx = local->fdctx; + + gf_proto_flock_to_flock (&rsp.flock, &lock); + + gf_log (this->name, GF_LOG_DEBUG, "%s type lock reacquired on file " + "with gfid %s from %"PRIu64 " to %"PRIu64, + get_lk_type (lock.l_type), uuid_utoa (fdctx->inode->gfid), + lock.l_start, lock.l_start + lock.l_len); + + if (clnt_fd_lk_local_unref (this, local) == 0 && + !clnt_fd_lk_local_error_status (this, local)) { + pthread_mutex_lock (&conf->lock); + { + fdctx->lk_heal_state = GF_LK_HEAL_DONE; + + list_add_tail (&fdctx->sfd_pos, + &conf->saved_fds); + } + pthread_mutex_unlock (&conf->lock); + + decrement_reopen_fd_count (this, conf); + } ret = 0; out: - if (ret < 0) + if (ret < 0) { clnt_fd_lk_local_mark_error (this, local); - (void) clnt_fd_lk_local_unref (this, local); + clnt_fd_lk_local_unref (this, local); + } + frame->local = NULL; STACK_DESTROY (frame->root); @@ -697,7 +792,10 @@ _client_reacquire_lock (xlator_t *this, clnt_fd_ctx_t *fdctx) local = clnt_fd_lk_local_create (fdctx); if (!local) { - clnt_release_reopen_fd (this, fdctx); + gf_log (this->name, GF_LOG_WARNING, "clnt_fd_lk_local_create " + "failed, aborting reacquring of locks on %s.", + uuid_utoa (fdctx->inode->gfid)); + clnt_reacquire_lock_error (this, fdctx, conf); goto out; } @@ -705,7 +803,9 @@ _client_reacquire_lock (xlator_t *this, clnt_fd_ctx_t *fdctx) memcpy (&flock, &fd_lk->user_flock, sizeof (struct gf_flock)); - ret = client_cmd_to_gf_cmd (fd_lk->cmd, &gf_cmd); + /* Always send F_SETLK even if the cmd was F_SETLKW */ + /* to avoid frame being blocked if lock cannot be granted. */ + ret = client_cmd_to_gf_cmd (F_SETLK, &gf_cmd); if (ret) { gf_log (this->name, GF_LOG_WARNING, "client_cmd_to_gf_cmd failed, " @@ -736,8 +836,12 @@ _client_reacquire_lock (xlator_t *this, clnt_fd_ctx_t *fdctx) client_reacquire_lock_cbk, NULL, NULL, 0, NULL, 0, NULL, (xdrproc_t)xdr_gfs3_lk_req); - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "reacquiring locks failed on file with gfid %s", + uuid_utoa (fdctx->inode->gfid)); break; + } ret = 0; frame = NULL; @@ -755,21 +859,24 @@ client_reacquire_lock (xlator_t *this, clnt_fd_ctx_t *fdctx) int32_t ret = -1; fd_lk_ctx_t *lk_ctx = NULL; + GF_VALIDATE_OR_GOTO ("client", this, out); + GF_VALIDATE_OR_GOTO (this->name, fdctx, out); + if (client_fd_lk_list_empty (fdctx->lk_ctx, _gf_false)) { - gf_log (this->name, GF_LOG_WARNING, + gf_log (this->name, GF_LOG_DEBUG, "fd lock list is empty"); - decrement_reopen_fd_count (this, (clnt_conf_t *)this->private); - ret = 0; - goto out; - } - - lk_ctx = fdctx->lk_ctx; + decrement_reopen_fd_count (this, + (clnt_conf_t *)this->private); + } else { + lk_ctx = fdctx->lk_ctx; - LOCK (&lk_ctx->lock); - { - ret = _client_reacquire_lock (this, fdctx); + LOCK (&lk_ctx->lock); + { + (void) _client_reacquire_lock (this, fdctx); + } + UNLOCK (&lk_ctx->lock); } - UNLOCK (&lk_ctx->lock); + ret = 0; out: return ret; } @@ -828,7 +935,6 @@ client3_1_reopen_cbk (struct rpc_req *req, struct iovec *iov, int count, } fdctx = local->fdctx; - if (!fdctx) { gf_log (frame->this->name, GF_LOG_WARNING, "fdctx not found"); ret = -1; @@ -839,9 +945,13 @@ client3_1_reopen_cbk (struct rpc_req *req, struct iovec *iov, int count, { fdctx->remote_fd = rsp.fd; if (!fdctx->released) { - list_add_tail (&fdctx->sfd_pos, &conf->saved_fds); - if (!client_fd_lk_list_empty (fdctx->lk_ctx, _gf_false)) + if (!client_fd_lk_list_empty (fdctx->lk_ctx, _gf_false)) { attempt_lock_recovery = _gf_true; + fdctx->lk_heal_state = GF_LK_HEAL_IN_PROGRESS; + } else { + list_add_tail (&fdctx->sfd_pos, + &conf->saved_fds); + } fdctx = NULL; } } @@ -850,21 +960,27 @@ client3_1_reopen_cbk (struct rpc_req *req, struct iovec *iov, int count, ret = 0; if (conf->lk_heal && attempt_lock_recovery) { - /* Delay decrement the reopen fd count untill all the + /* Delay decrementing the reopen fd count untill all the locks corresponding to this fd are acquired.*/ - gf_log (frame->this->name, GF_LOG_WARNING, "acquiring locks on " - "%s", local->loc.path); + gf_log (this->name, GF_LOG_DEBUG, "acquiring locks " + "on %s", local->loc.path); ret = client_reacquire_lock (frame->this, local->fdctx); + if (ret) { + clnt_reacquire_lock_error (this, local->fdctx, conf); + gf_log (this->name, GF_LOG_WARNING, "acquiring locks " + "failed on %s", local->loc.path); + ret = 0; + } } else { fd_count = decrement_reopen_fd_count (frame->this, conf); } -out: - if (fdctx) - client_fdctx_destroy (this, fdctx); - - if ((ret < 0) && frame && frame->this && conf) +out: + if (fdctx) { + clnt_release_reopen_fd (this, fdctx); + } else if ((ret < 0) && frame && frame->this && conf) { decrement_reopen_fd_count (frame->this, conf); + } if (frame) { frame->local = NULL; @@ -1172,7 +1288,7 @@ client_post_handshake (call_frame_t *frame, xlator_t *this) } } else { gf_log (this->name, GF_LOG_DEBUG, - "no fds to open - notifying all parents child up"); + "No fds to open - notifying all parents child up"); client_set_lk_version (this); client_notify_parents_child_up (this); } @@ -1284,7 +1400,7 @@ client_setvolume_cbk (struct rpc_req *req, struct iovec *iov, int count, void *m goto out; } - gf_log (this->name, GF_LOG_INFO, "clnt-lk-version = %d, " + gf_log (this->name, GF_LOG_DEBUG, "clnt-lk-version = %d, " "server-lk-version = %d", client_get_lk_ver (conf), lk_ver); /* TODO: currently setpeer path is broken */ /* @@ -1322,13 +1438,17 @@ client_setvolume_cbk (struct rpc_req *req, struct iovec *iov, int count, void *m conf->need_different_port = 0; if (lk_ver != client_get_lk_ver (conf)) { + gf_log (this->name, GF_LOG_INFO, "Server and Client " + "lk-version numbers are not same, reopening the fds"); client_mark_fd_bad (this); client_post_handshake (frame, frame->this); } else { /*TODO: Traverse the saved fd list, and send release to the server on fd's that were closed during grace period */ - ; + gf_log (this->name, GF_LOG_INFO, "Server and Client " + "lk-version numbers are same, no need to " + "reopen the fds"); } out: |