summaryrefslogtreecommitdiffstats
path: root/xlators/cluster/afr/src/afr.c
diff options
context:
space:
mode:
Diffstat (limited to 'xlators/cluster/afr/src/afr.c')
-rw-r--r--xlators/cluster/afr/src/afr.c217
1 files changed, 200 insertions, 17 deletions
diff --git a/xlators/cluster/afr/src/afr.c b/xlators/cluster/afr/src/afr.c
index 8e94d5497..c724eb2ae 100644
--- a/xlators/cluster/afr/src/afr.c
+++ b/xlators/cluster/afr/src/afr.c
@@ -107,6 +107,7 @@ reconfigure (xlator_t *this, dict_t *options)
{
afr_private_t *priv = NULL;
xlator_t *read_subvol = NULL;
+ int read_subvol_index = -1;
int ret = -1;
int index = -1;
char *qtype = NULL;
@@ -149,6 +150,9 @@ reconfigure (xlator_t *this, dict_t *options)
GF_OPTION_RECONF ("read-subvolume", read_subvol, options, xlator, out);
+ GF_OPTION_RECONF ("read-hash-mode", priv->hash_mode,
+ options, uint32, out);
+
if (read_subvol) {
index = xlator_subvolume_index (this, read_subvol);
if (index == -1) {
@@ -159,6 +163,18 @@ reconfigure (xlator_t *this, dict_t *options)
priv->read_child = index;
}
+ GF_OPTION_RECONF ("read-subvolume-index",read_subvol_index, options,int32,out);
+
+ if (read_subvol_index >-1) {
+ index=read_subvol_index;
+ if (index >= priv->child_count) {
+ gf_log (this->name, GF_LOG_ERROR, "%d not a subvolume-index",
+ index);
+ goto out;
+ }
+ priv->read_child = index;
+ }
+
GF_OPTION_RECONF ("eager-lock", priv->eager_lock, options, bool, out);
GF_OPTION_RECONF ("quorum-type", qtype, options, str, out);
GF_OPTION_RECONF ("quorum-count", priv->quorum_count, options,
@@ -167,6 +183,18 @@ reconfigure (xlator_t *this, dict_t *options)
GF_OPTION_RECONF ("heal-timeout", priv->shd.timeout, options,
int32, out);
+ GF_OPTION_RECONF ("post-op-delay-secs", priv->post_op_delay_secs, options,
+ uint32, out);
+
+ GF_OPTION_RECONF (AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size,
+ options, size, out);
+ /* Reset this so we re-discover in case the topology changed. */
+ GF_OPTION_RECONF ("readdir-failover", priv->readdir_failover, options,
+ bool, out);
+ GF_OPTION_RECONF ("ensure-durability", priv->ensure_durability, options,
+ bool, out);
+ priv->did_discovery = _gf_false;
+
ret = 0;
out:
return ret;
@@ -193,6 +221,7 @@ init (xlator_t *this)
int ret = -1;
GF_UNUSED int op_errno = 0;
xlator_t *read_subvol = NULL;
+ int read_subvol_index = -1;
xlator_t *fav_child = NULL;
char *qtype = NULL;
@@ -224,7 +253,6 @@ init (xlator_t *this)
priv->child_count = child_count;
-
priv->read_child = -1;
GF_OPTION_INIT ("read-subvolume", read_subvol, xlator, out);
@@ -236,6 +264,18 @@ init (xlator_t *this)
goto out;
}
}
+ GF_OPTION_INIT ("read-subvolume-index",read_subvol_index,int32,out);
+ if (read_subvol_index > -1) {
+ if (read_subvol_index >= priv->child_count) {
+ gf_log (this->name, GF_LOG_ERROR, "%d not a subvolume-index",
+ read_subvol_index);
+ goto out;
+ }
+ priv->read_child = read_subvol_index;
+ }
+ GF_OPTION_INIT ("choose-local", priv->choose_local, bool, out);
+
+ GF_OPTION_INIT ("read-hash-mode", priv->hash_mode, uint32, out);
priv->favorite_child = -1;
GF_OPTION_INIT ("favorite-child", fav_child, xlator, out);
@@ -291,8 +331,15 @@ init (xlator_t *this)
GF_OPTION_INIT ("eager-lock", priv->eager_lock, bool, out);
GF_OPTION_INIT ("quorum-type", qtype, str, out);
GF_OPTION_INIT ("quorum-count", priv->quorum_count, uint32, out);
+ GF_OPTION_INIT (AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size, size,
+ out);
fix_quorum_options(this,priv,qtype);
+ GF_OPTION_INIT ("post-op-delay-secs", priv->post_op_delay_secs, uint32, out);
+ GF_OPTION_INIT ("readdir-failover", priv->readdir_failover, bool, out);
+ GF_OPTION_INIT ("ensure-durability", priv->ensure_durability, bool,
+ out);
+
priv->wait_count = 1;
priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
@@ -333,8 +380,6 @@ init (xlator_t *this)
AFR_XATTR_PREFIX,
trav->xlator->name);
if (-1 == ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "asprintf failed to set pending key");
ret = -ENOMEM;
goto out;
}
@@ -343,6 +388,13 @@ init (xlator_t *this)
i++;
}
+ ret = gf_asprintf (&priv->sh_domain, AFR_SH_DATA_DOMAIN_FMT,
+ this->name);
+ if (-1 == ret) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
priv->last_event = GF_CALLOC (child_count, sizeof (*priv->last_event),
gf_afr_mt_int32_t);
if (!priv->last_event) {
@@ -387,15 +439,18 @@ init (xlator_t *this)
if (!priv->shd.timer)
goto out;
- priv->shd.healed = eh_new (AFR_EH_HEALED_LIMIT, _gf_false);
+ priv->shd.healed = eh_new (AFR_EH_HEALED_LIMIT, _gf_false,
+ _destroy_shd_event_data);
if (!priv->shd.healed)
goto out;
- priv->shd.heal_failed = eh_new (AFR_EH_HEAL_FAIL_LIMIT, _gf_false);
+ priv->shd.heal_failed = eh_new (AFR_EH_HEAL_FAIL_LIMIT, _gf_false,
+ _destroy_shd_event_data);
if (!priv->shd.heal_failed)
goto out;
- priv->shd.split_brain = eh_new (AFR_EH_SPLIT_BRAIN_LIMIT, _gf_false);
+ priv->shd.split_brain = eh_new (AFR_EH_SPLIT_BRAIN_LIMIT, _gf_false,
+ _destroy_shd_event_data);
if (!priv->shd.split_brain)
goto out;
@@ -405,7 +460,9 @@ init (xlator_t *this)
priv->root_inode = inode_ref (this->itable->root);
GF_OPTION_INIT ("node-uuid", priv->shd.node_uuid, str, out);
GF_OPTION_INIT ("heal-timeout", priv->shd.timeout, int32, out);
-
+ ret = afr_initialise_statistics (this);
+ if (ret)
+ goto out;
ret = 0;
out:
return ret;
@@ -440,6 +497,9 @@ struct xlator_fops fops = {
.finodelk = afr_finodelk,
.entrylk = afr_entrylk,
.fentrylk = afr_fentrylk,
+ .fallocate = afr_fallocate,
+ .discard = afr_discard,
+ .zerofill = afr_zerofill,
/* inode read */
.access = afr_access,
@@ -492,33 +552,79 @@ struct xlator_cbks cbks = {
struct volume_options options[] = {
{ .key = {"read-subvolume" },
- .type = GF_OPTION_TYPE_XLATOR
+ .type = GF_OPTION_TYPE_XLATOR,
+ .description = "inode-read fops happen only on one of the bricks in "
+ "replicate. Afr will prefer the one specified using "
+ "this option if it is not stale. Option value must be "
+ "one of the xlator names of the children. "
+ "Ex: <volname>-client-0 till "
+ "<volname>-client-<number-of-bricks - 1>"
+ },
+ { .key = {"read-subvolume-index" },
+ .type = GF_OPTION_TYPE_INT,
+ .default_value = "-1",
+ .description = "inode-read fops happen only on one of the bricks in "
+ "replicate. AFR will prefer the one specified using "
+ "this option if it is not stale. allowed options"
+ " include -1 till replica-count - 1"
+ },
+ { .key = {"read-hash-mode" },
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = 2,
+ .default_value = "0",
+ .description = "inode-read fops happen only on one of the bricks in "
+ "replicate. AFR will prefer the one computed using "
+ "the method specified using this option"
+ "0 = first responder, "
+ "1 = hash by GFID of file (all clients use "
+ "same subvolume), "
+ "2 = hash by GFID of file and client PID",
+ },
+ { .key = {"choose-local" },
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "true",
+ .description = "Choose a local subvolume(i.e. Brick) to read from if "
+ "read-subvolume is not explicitly set.",
},
{ .key = {"favorite-child"},
- .type = GF_OPTION_TYPE_XLATOR
+ .type = GF_OPTION_TYPE_XLATOR,
+ .description = "If a split-brain happens choose subvol/brick set by "
+ "this option as source."
},
{ .key = {"background-self-heal-count"},
.type = GF_OPTION_TYPE_INT,
.min = 0,
.default_value = "16",
+ .validate = GF_OPT_VALIDATE_MIN,
+ .description = "This specifies the number of self-heals that can be "
+ " performed in background without blocking the fop"
},
{ .key = {"data-self-heal"},
.type = GF_OPTION_TYPE_STR,
- .default_value = "",
.value = {"1", "on", "yes", "true", "enable",
"0", "off", "no", "false", "disable",
"open"},
.default_value = "on",
+ .description = "Using this option we can enable/disable data "
+ "self-heal on the file. \"open\" means data "
+ "self-heal action will only be triggered by file "
+ "open operations."
},
{ .key = {"data-self-heal-algorithm"},
.type = GF_OPTION_TYPE_STR,
- .default_value = "",
.description = "Select between \"full\", \"diff\". The "
"\"full\" algorithm copies the entire file from "
"source to sink. The \"diff\" algorithm copies to "
"sink only those blocks whose checksums don't match "
- "with those of source.",
- .value = { "diff", "full", "" }
+ "with those of source. If no option is configured "
+ "the option is chosen dynamically as follows: "
+ "If the file does not exist on one of the sinks "
+ "or empty file exists or if the source file size is "
+ "about the same as page size the entire file will "
+ "be read and written i.e \"full\" algo, "
+ "otherwise \"diff\" algo is chosen.",
+ .value = { "diff", "full"}
},
{ .key = {"data-self-heal-window-size"},
.type = GF_OPTION_TYPE_INT,
@@ -531,26 +637,43 @@ struct volume_options options[] = {
{ .key = {"metadata-self-heal"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "on",
+ .description = "Using this option we can enable/disable metadata "
+ "i.e. Permissions, ownerships, xattrs self-heal on "
+ "the file/directory."
},
{ .key = {"entry-self-heal"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "on",
+ .description = "Using this option we can enable/disable entry "
+ "self-heal on the directory."
},
{ .key = {"data-change-log"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "on",
+ .description = "Data fops like write/truncate will not perform "
+ "pre/post fop changelog operations in afr transaction "
+ "if this option is disabled"
},
{ .key = {"metadata-change-log"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "on",
+ .description = "Metadata fops like setattr/setxattr will not perform "
+ "pre/post fop changelog operations in afr transaction "
+ "if this option is disabled"
},
{ .key = {"entry-change-log"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "on",
+ .description = "Entry fops like create/unlink will not perform "
+ "pre/post fop changelog operations in afr transaction "
+ "if this option is disabled"
},
{ .key = {"optimistic-change-log"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "on",
+ .description = "Entry/Metadata fops will not perform "
+ "pre fop changelog operations in afr transaction "
+ "if this option is enabled."
},
{ .key = {"strict-readdir"},
.type = GF_OPTION_TYPE_BOOL,
@@ -559,26 +682,55 @@ struct volume_options options[] = {
{ .key = {"inodelk-trace"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "off",
+ .description = "Enabling this option logs inode lock/unlocks"
},
{ .key = {"entrylk-trace"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "off",
+ .description = "Enabling this option logs entry lock/unlocks"
},
{ .key = {"eager-lock"},
.type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
+ .default_value = "on",
+ .description = "Lock phase of a transaction has two sub-phases. "
+ "First is an attempt to acquire locks in parallel by "
+ "broadcasting non-blocking lock requests. If lock "
+ "aquistion fails on any server, then the held locks "
+ "are unlocked and revert to a blocking locked mode "
+ "sequentially on one server after another. If this "
+ "option is enabled the initial broadcasting lock "
+ "request attempt to acquire lock on the entire file. "
+ "If this fails, we revert back to the sequential "
+ "\"regional\" blocking lock as before. In the case "
+ "where such an \"eager\" lock is granted in the "
+ "non-blocking phase, it gives rise to an opportunity "
+ "for optimization. i.e, if the next write transaction "
+ "on the same FD arrives before the unlock phase of "
+ "the first transaction, it \"takes over\" the full "
+ "file lock. Similarly if yet another data transaction "
+ "arrives before the unlock phase of the \"optimized\" "
+ "transaction, that in turn \"takes over\" the lock as "
+ "well. The actual unlock now happens at the end of "
+ "the last \"optimzed\" transaction."
+
},
{ .key = {"self-heal-daemon"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "off",
+ .description = "This option applies to only self-heal-daemon. "
+ "Index directory crawl and automatic healing of files"
+ "will not be performed if this option is turned off."
},
{ .key = {"iam-self-heal-daemon"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "off",
+ .description = "This option differentiates if the replicate "
+ "translator is running as part of self-heal-daemon "
+ "or not."
},
{ .key = {"quorum-type"},
.type = GF_OPTION_TYPE_STR,
- .value = { "none", "auto", "fixed", "" },
+ .value = { "none", "auto", "fixed"},
.default_value = "none",
.description = "If value is \"fixed\" only allow writes if "
"quorum-count bricks are present. If value is "
@@ -597,14 +749,45 @@ struct volume_options options[] = {
},
{ .key = {"node-uuid"},
.type = GF_OPTION_TYPE_STR,
- .description = "Local glusterd uuid string",
+ .description = "Local glusterd uuid string, used in starting "
+ "self-heal-daemon so that it can crawl only on "
+ "local index directories.",
},
{ .key = {"heal-timeout"},
.type = GF_OPTION_TYPE_INT,
.min = 60,
.max = INT_MAX,
.default_value = "600",
- .description = "Poll timeout for checking the need to self-heal"
+ .description = "time interval for checking the need to self-heal "
+ "in self-heal-daemon"
+ },
+ { .key = {"post-op-delay-secs"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = INT_MAX,
+ .default_value = "1",
+ .description = "Time interval induced artificially before "
+ "post-operation phase of the transaction to "
+ "enhance overlap of adjacent write operations.",
+ },
+ { .key = {AFR_SH_READDIR_SIZE_KEY},
+ .type = GF_OPTION_TYPE_SIZET,
+ .description = "readdirp size for performing entry self-heal",
+ .min = 1024,
+ .max = 131072,
+ .default_value = "1KB",
+ },
+ { .key = {"readdir-failover"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .description = "readdir(p) will not failover if this option is off",
+ .default_value = "on",
+ },
+ { .key = {"ensure-durability"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .description = "Afr performs fsyncs for transactions if this "
+ "option is on to make sure the changelogs/data is "
+ "written to the disk",
+ .default_value = "on",
},
{ .key = {NULL} },
};
generated by cgit (git 2.34.1) at 2026-01-05 03:29:00 +0000