diff options
author | Ravishankar N <ravishankar@redhat.com> | 2017-09-20 12:16:06 +0530 |
---|---|---|
committer | Pranith Kumar Karampuri <pkarampu@redhat.com> | 2017-10-09 06:23:08 +0000 |
commit | 20fa80057eb430fd72b4fa31b9b65598b8ec1265 (patch) | |
tree | 3e1dd8c0cb7a394ebcdc7004ccaddb935c67e4d6 /xlators/cluster/afr/src/afr-self-heal-name.c | |
parent | 938addeb7ec634e431c2c8c0a768a2a9ed056c0d (diff) |
afr: heal gfid as a part of entry heal
Problem:
If a brick crashes after an entry (file or dir) is created but before
gfid is assigned, the good bricks will have pending entry heal xattrs
but the heal won't complete because afr_selfheal_recreate_entry() tries
to create the entry again and it fails with EEXIST.
Fix:
We could have fixed posx_mknod/mkdir etc to assign the gfid if the file
already exists but the right thing to do seems to be to trigger a lookup
on the bad brick and let it heal the gfid instead of winding an
mknod/mkdir in the first place.
Change-Id: I82f76665a7541f1893ef8d847b78af6466aff1ff
BUG: 1493415
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
Diffstat (limited to 'xlators/cluster/afr/src/afr-self-heal-name.c')
-rw-r--r-- | xlators/cluster/afr/src/afr-self-heal-name.c | 64 |
1 files changed, 4 insertions, 60 deletions
diff --git a/xlators/cluster/afr/src/afr-self-heal-name.c b/xlators/cluster/afr/src/afr-self-heal-name.c index 2aad4c75bee..352d151207e 100644 --- a/xlators/cluster/afr/src/afr-self-heal-name.c +++ b/xlators/cluster/afr/src/afr-self-heal-name.c @@ -18,50 +18,18 @@ int __afr_selfheal_assign_gfid (xlator_t *this, inode_t *parent, uuid_t pargfid, const char *bname, inode_t *inode, struct afr_reply *replies, void *gfid, - unsigned char *locked_on, + unsigned char *locked_on, int source, gf_boolean_t is_gfid_absent) { int ret = 0; int up_count = 0; int locked_count = 0; - int i = 0; afr_private_t *priv = NULL; - dict_t *xdata = NULL; - loc_t loc = {0, }; - call_frame_t *new_frame = NULL; - afr_local_t *new_local = NULL; - unsigned char *wind_on = NULL; priv = this->private; - wind_on = alloca0 (priv->child_count); - - new_frame = afr_frame_create (this); - if (!new_frame) { - ret = -ENOMEM; - goto out; - } - - new_local = new_frame->local; gf_uuid_copy (parent->gfid, pargfid); - xdata = dict_new (); - if (!xdata) { - ret = -ENOMEM; - goto out; - } - - ret = dict_set_static_bin (xdata, "gfid-req", gfid, 16); - if (ret) { - ret = -ENOMEM; - goto out; - } - - loc.parent = inode_ref (parent); - loc.inode = inode_ref (inode); - gf_uuid_copy (loc.pargfid, pargfid); - loc.name = bname; - if (is_gfid_absent) { /* Ensure all children of AFR are up before performing gfid heal, to * guard against the possibility of gfid split brain. */ @@ -79,34 +47,10 @@ __afr_selfheal_assign_gfid (xlator_t *this, inode_t *parent, uuid_t pargfid, } } - /* gfid heal on those subvolumes that do not have gfid associated - * with the inode and update those replies. - */ - for (i = 0; i < priv->child_count; i++) { - if (replies[i].valid && replies[i].op_ret == 0 && - !gf_uuid_is_null (replies[i].poststat.ia_gfid) && - !gf_uuid_compare (replies[i].poststat.ia_gfid, gfid)) - continue; - wind_on[i] = 1; - } - - AFR_ONLIST (wind_on, new_frame, afr_selfheal_discover_cbk, lookup, - &loc, xdata); - - for (i = 0; i < priv->child_count; i++) { - if (!wind_on[i]) - continue; - afr_reply_wipe (&replies[i]); - afr_reply_copy (&replies[i], &new_local->replies[i]); - } + afr_lookup_and_heal_gfid (this, parent, bname, inode, replies, source, + gfid); out: - loc_wipe (&loc); - if (xdata) - dict_unref (xdata); - if (new_frame) - AFR_STACK_DESTROY (new_frame); - return ret; } @@ -481,7 +425,7 @@ __afr_selfheal_name_do (call_frame_t *frame, xlator_t *this, inode_t *parent, is_gfid_absent = (gfid_idx == -1) ? _gf_true : _gf_false; ret = __afr_selfheal_assign_gfid (this, parent, pargfid, bname, inode, - replies, gfid, locked_on, + replies, gfid, locked_on, source, is_gfid_absent); if (ret) return ret; |