From 2588152f410c133c408d182f04b8cc5759c1af3e Mon Sep 17 00:00:00 2001 From: Ravishankar N Date: Wed, 10 May 2017 10:03:08 +0530 Subject: [PATCH 437/473] afr: propagate correct errno for fop failures in arbiter Backport of: https://review.gluster.org/17235 Problem: If quorum is not met in fop cbk, arbiter sends an ENOTCONN error to the upper xlators. In a VM workload with sharding enabled, this was leading to the VM pausing when replace-brick was performed as described in the BZ. Fix: Move the fop cbk arbitration logic to afr_handle_quorum() because in normal replica volumes, that is the function that has the quorum and errno checks in the fop cbk path before doing a post-op. Thanks to Pranith for suggesting this approach. Change-Id: I4fffa3b34a4017eb950b4e1c55b943441eacee21 BUG: 1443980 Signed-off-by: Ravishankar N Reviewed-on: https://code.engineering.redhat.com/gerrit/106202 --- xlators/cluster/afr/src/afr-dir-write.c | 1 - xlators/cluster/afr/src/afr-inode-write.c | 1 - xlators/cluster/afr/src/afr-transaction.c | 23 ++++++++++++----------- xlators/cluster/afr/src/afr-transaction.h | 2 -- 4 files changed, 12 insertions(+), 15 deletions(-) diff --git a/xlators/cluster/afr/src/afr-dir-write.c b/xlators/cluster/afr/src/afr-dir-write.c index 8e483c3..9099b8c 100644 --- a/xlators/cluster/afr/src/afr-dir-write.c +++ b/xlators/cluster/afr/src/afr-dir-write.c @@ -183,7 +183,6 @@ __afr_dir_write_finalize (call_frame_t *frame, xlator_t *this) } } - afr_txn_arbitrate_fop_cbk (frame, this); } diff --git a/xlators/cluster/afr/src/afr-inode-write.c b/xlators/cluster/afr/src/afr-inode-write.c index 200b420..04bbf21 100644 --- a/xlators/cluster/afr/src/afr-inode-write.c +++ b/xlators/cluster/afr/src/afr-inode-write.c @@ -131,7 +131,6 @@ __afr_inode_write_finalize (call_frame_t *frame, xlator_t *this) } } - afr_txn_arbitrate_fop_cbk (frame, this); afr_set_in_flight_sb_status (this, local, local->inode); } diff --git a/xlators/cluster/afr/src/afr-transaction.c b/xlators/cluster/afr/src/afr-transaction.c index a3cb766..644ebe2 100644 --- a/xlators/cluster/afr/src/afr-transaction.c +++ b/xlators/cluster/afr/src/afr-transaction.c @@ -303,22 +303,21 @@ afr_compute_pre_op_sources (call_frame_t *frame, xlator_t *this) } } -void -afr_txn_arbitrate_fop_cbk (call_frame_t *frame, xlator_t *this) +gf_boolean_t +afr_has_arbiter_fop_cbk_quorum (call_frame_t *frame) { afr_local_t *local = NULL; afr_private_t *priv = NULL; + xlator_t *this = NULL; gf_boolean_t fop_failed = _gf_false; unsigned char *pre_op_sources = NULL; int i = 0; local = frame->local; + this = frame->this; priv = this->private; pre_op_sources = local->transaction.pre_op_sources; - if (priv->arbiter_count != 1 || local->op_ret < 0) - return; - /* If the fop failed on the brick, it is not a source. */ for (i = 0; i < priv->child_count; i++) if (local->transaction.failed_subvols[i]) @@ -334,12 +333,10 @@ afr_txn_arbitrate_fop_cbk (call_frame_t *frame, xlator_t *this) break; } - if (fop_failed) { - local->op_ret = -1; - local->op_errno = ENOTCONN; - } + if (fop_failed) + return _gf_false; - return; + return _gf_true; } void @@ -787,8 +784,12 @@ afr_handle_quorum (call_frame_t *frame) * no split-brain with the fix. The problem is eliminated completely. */ - if (afr_has_fop_cbk_quorum (frame)) + if (priv->arbiter_count) { + if (afr_has_arbiter_fop_cbk_quorum (frame)) + return; + } else if (afr_has_fop_cbk_quorum (frame)) { return; + } for (i = 0; i < priv->child_count; i++) { if (local->transaction.pre_op[i]) diff --git a/xlators/cluster/afr/src/afr-transaction.h b/xlators/cluster/afr/src/afr-transaction.h index db82456..dd19e5b 100644 --- a/xlators/cluster/afr/src/afr-transaction.h +++ b/xlators/cluster/afr/src/afr-transaction.h @@ -16,8 +16,6 @@ void afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this, int child_index); -void -afr_txn_arbitrate_fop_cbk (call_frame_t *frame, xlator_t *this); int afr_lock_server_count (afr_private_t *priv, afr_transaction_type type); -- 1.8.3.1