diff options
| author | Ravishankar N <ravishankar@redhat.com> | 2017-01-30 09:54:16 +0530 |
|---|---|---|
| committer | Pranith Kumar Karampuri <pkarampu@redhat.com> | 2017-02-09 20:37:00 -0500 |
| commit | 0e03336a9362e5717e561f76b0c543e5a197b31b (patch) | |
| tree | 28be72abb87f19d28b82c9ac7cec61533a968d05 /xlators/cluster/afr/src/afr-self-heal-common.c | |
| parent | 55625293093d485623f3f3d98687cd1e2c594460 (diff) | |
afr: all children of AFR must be up to resolve s-brain
Problem:
The various split-brain resolution policies (favorite-child-policy based,
CLI based and mount (get/setfattr) based) attempt to resolve split-brain
even when not all bricks of replica are up. This can be a problem when
say in a replica 3, the only good copy is down and the other 2 bricks
are up and blame each other (i.e. split-brain). We end up healing the
file in such a case and allow I/O on it.
Fix:
A decision on whether the file is in split-brain or not must be taken
only if we are able to examine the afr xattrs of *all* bricks of a given
replica.
Change-Id: Icddb1268b380005799990f5379ef957d84639ef9
BUG: 1417522
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
Reviewed-on: https://review.gluster.org/16476
Smoke: Gluster Build System <jenkins@build.gluster.org>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>
Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
Diffstat (limited to 'xlators/cluster/afr/src/afr-self-heal-common.c')
| -rw-r--r-- | xlators/cluster/afr/src/afr-self-heal-common.c | 38 |
1 files changed, 34 insertions, 4 deletions
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c index 17e15d760c6..a8d3febb21b 100644 --- a/xlators/cluster/afr/src/afr-self-heal-common.c +++ b/xlators/cluster/afr/src/afr-self-heal-common.c @@ -473,6 +473,19 @@ afr_dict_contains_heal_op (call_frame_t *frame) return _gf_true; } +gf_boolean_t +afr_can_decide_split_brain_source_sinks (struct afr_reply *replies, + int child_count) +{ + int i = 0; + + for (i = 0; i < child_count; i++) + if (replies[i].valid != 1 || replies[i].op_ret != 0) + return _gf_false; + + return _gf_true; +} + int afr_mark_split_brain_source_sinks_by_heal_op (call_frame_t *frame, xlator_t *this, unsigned char *sources, @@ -511,6 +524,14 @@ afr_mark_split_brain_source_sinks_by_heal_op (call_frame_t *frame, } xdata_rsp = local->xdata_rsp; + if (!afr_can_decide_split_brain_source_sinks (replies, + priv->child_count)) { + ret = dict_set_str (xdata_rsp, "sh-fail-msg", + SBRAIN_HEAL_NO_GO_MSG); + ret = -1; + goto out; + } + for (i = 0 ; i < priv->child_count; i++) if (locked_on[i]) sources[i] = 1; @@ -749,26 +770,35 @@ afr_sh_get_fav_by_policy (xlator_t *this, struct afr_reply *replies, int fav_child = -1; priv = this->private; + if (!afr_can_decide_split_brain_source_sinks (replies, + priv->child_count)) { + return -1; + } + switch (priv->fav_child_policy) { case AFR_FAV_CHILD_BY_SIZE: fav_child = afr_sh_fav_by_size (this, replies, inode); - if (policy_str && fav_child >= 0) + if (policy_str && fav_child >= 0) { *policy_str = "SIZE"; + } break; case AFR_FAV_CHILD_BY_CTIME: fav_child = afr_sh_fav_by_ctime (this, replies, inode); - if (policy_str && fav_child >= 0) + if (policy_str && fav_child >= 0) { *policy_str = "CTIME"; + } break; case AFR_FAV_CHILD_BY_MTIME: fav_child = afr_sh_fav_by_mtime (this, replies, inode); - if (policy_str && fav_child >= 0) + if (policy_str && fav_child >= 0) { *policy_str = "MTIME"; + } break; case AFR_FAV_CHILD_BY_MAJORITY: fav_child = afr_sh_fav_by_majority (this, replies, inode); - if (policy_str && fav_child >= 0) + if (policy_str && fav_child >= 0) { *policy_str = "MAJORITY"; + } break; case AFR_FAV_CHILD_NONE: default: |