summaryrefslogtreecommitdiffstats
path: root/xlators/cluster
diff options
context:
space:
mode:
authorEmmanuel Dreyfus <manu@netbsd.org>2014-12-23 06:54:50 +0100
committerPranith Kumar Karampuri <pkarampu@redhat.com>2014-12-22 23:29:54 -0800
commit9a6ea92df1007cc6f402ff7cc560e3206fa42e62 (patch)
treefe31bb641e4e4c3684e81fe14a71534636d0593d /xlators/cluster
parentda9deb54df91dedc51ebe165f3a0be646455cb5b (diff)
Avoid spurious directory metedata split brain
When directory content is modified, [mc]time is updated. On Linux, the filesystem does it, while at least on NetBSD, the kernel file-system independant code does it. This means that when entries are added while bricks are down, the kernel sends a SETATTR [mc]time which will cause metadata split brain for the directory. In this case, clear the split brain by finding the source with the most recent modification date. BUG: 1129939 Change-Id: Ic0177e0df753a4748624d0b906834ed54593adb9 Signed-off-by: Emmanuel Dreyfus <manu@netbsd.org> Reviewed-on: http://review.gluster.org/9291 Reviewed-by: Ravishankar N <ravishankar@redhat.com> Reviewed-by: Krutika Dhananjay <kdhananj@redhat.com> Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com> Tested-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
Diffstat (limited to 'xlators/cluster')
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-metadata.c116
1 files changed, 115 insertions, 1 deletions
diff --git a/xlators/cluster/afr/src/afr-self-heal-metadata.c b/xlators/cluster/afr/src/afr-self-heal-metadata.c
index 68dc568bc65..87600df3bad 100644
--- a/xlators/cluster/afr/src/afr-self-heal-metadata.c
+++ b/xlators/cluster/afr/src/afr-self-heal-metadata.c
@@ -99,6 +99,96 @@ out:
return ret;
}
+static inline uint64_t
+mtime_ns(struct iatt *ia)
+{
+ uint64_t ret;
+
+ ret = (((uint64_t)(ia->ia_mtime)) * 1000000000)
+ + (uint64_t)(ia->ia_mtime_nsec);
+
+ return ret;
+}
+
+/*
+ * When directory content is modified, [mc]time is updated. On
+ * Linux, the filesystem does it, while at least on NetBSD, the
+ * kernel file-system independant code does it. This means that
+ * when entries are added while bricks are down, the kernel sends
+ * a SETATTR [mc]time which will cause metadata split brain for
+ * the directory. In this case, clear the split brain by finding
+ * the source with the most recent modification date.
+ */
+static int
+afr_dirtime_splitbrain_source (call_frame_t *frame, xlator_t *this,
+ struct afr_reply *replies,
+ unsigned char *locked_on)
+{
+ afr_private_t *priv = NULL;
+ int source = -1;
+ struct iatt source_ia;
+ struct iatt child_ia;
+ uint64_t mtime = 0;
+ int i;
+ int ret = -1;
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!locked_on[i])
+ continue;
+
+ if (!replies[i].valid)
+ continue;
+
+ if (replies[i].op_ret != 0)
+ continue;
+
+ if (mtime_ns(&replies[i].poststat) <= mtime)
+ continue;
+
+ mtime = mtime_ns(&replies[i].poststat);
+ source = i;
+ }
+
+ if (source == -1)
+ goto out;
+
+ source_ia = replies[source].poststat;
+ if (source_ia.ia_type != IA_IFDIR)
+ goto out;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source)
+ continue;
+
+ if (!replies[i].valid)
+ continue;
+
+ if (replies[i].op_ret != 0)
+ continue;
+
+ child_ia = replies[i].poststat;
+
+ if (!IA_EQUAL(source_ia, child_ia, gfid) ||
+ !IA_EQUAL(source_ia, child_ia, type) ||
+ !IA_EQUAL(source_ia, child_ia, prot) ||
+ !IA_EQUAL(source_ia, child_ia, uid) ||
+ !IA_EQUAL(source_ia, child_ia, gid) ||
+ !afr_xattrs_are_equal (replies[source].xdata,
+ replies[i].xdata))
+ goto out;
+ }
+
+ /*
+ * Metadata split brain is just about [amc]time
+ * We return our source.
+ */
+ ret = source;
+out:
+ return ret;
+}
+
/*
* Look for mismatching uid/gid or mode or user xattrs even if
@@ -122,10 +212,34 @@ __afr_selfheal_metadata_finalize_source (call_frame_t *frame, xlator_t *this,
sources_count = AFR_COUNT (sources, priv->child_count);
if ((AFR_CMP (locked_on, healed_sinks, priv->child_count) == 0)
- || !sources_count) {
+ || !sources_count) {
+ /* If this is a directory mtime/ctime only split brain
+ use the most recent */
+ source = afr_dirtime_splitbrain_source (frame, this,
+ replies, locked_on);
+ if (source != -1) {
+ gf_log (this->name, GF_LOG_NOTICE, "clear time "
+ "split brain on %s",
+ uuid_utoa (replies[source].poststat.ia_gfid));
+ sources[source] = 1;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source)
+ continue;
+
+ if (!locked_on[i])
+ continue;
+
+ healed_sinks[i] = 1;
+ }
+
+ return source;
+ }
+
if (!priv->metadata_splitbrain_forced_heal) {
return -EIO;
}
+
/* Metadata split brain, select one subvol
arbitrarily */
for (i = 0; i < priv->child_count; i++) {