fluencelabs/redis
1
0
Fork 0
mirror of https://github.com/fluencelabs/redis synced 2025-03-19 09:00:51 +00:00
Code Issues Projects Releases Wiki Activity

Cluster: refactoring of slots update code and more.

Browse Source 
The commit also introduces detection of nodes publishing not updated
configuration. More work in progress to send an UPDATE packet to inform
of the config change.
This commit is contained in:
antirez 2013-11-08 10:32:16 +01:00
parent d3588dc194
commit 6c6572be95
1 changed files with 98 additions and 54 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

152
src/cluster.c
View File

@ -1029,7 +1029,7 @@ int clusterProcessPacket(clusterLink *link) {
} }
} }
/* Update master/slave state */ /* Check for role switch: slave -> master or master -> slave. */
if (sender) { if (sender) {
if (!memcmp(hdr->slaveof,REDIS_NODE_NULL_NAME, if (!memcmp(hdr->slaveof,REDIS_NODE_NULL_NAME,
sizeof(hdr->slaveof))) sizeof(hdr->slaveof)))
@ -1079,67 +1079,111 @@ int clusterProcessPacket(clusterLink *link) {
} }
/* Update our info about served slots. /* Update our info about served slots.
*
* Note: this MUST happen after we update the master/slave state * Note: this MUST happen after we update the master/slave state
* so that REDIS_NODE_MASTER flag will be set. */ * so that REDIS_NODE_MASTER flag will be set. */
if (sender && sender->flags & REDIS_NODE_MASTER) {
int changes, j;
changes = /* Many checks are only needed if the set of served slots this
memcmp(sender->slots,hdr->myslots,sizeof(hdr->myslots)) != 0; * instance claims is different compared to the set of slots we have for
if (changes) { * it. Check this ASAP to avoid other computational expansive checks later. */
clusterNode *curmaster, *newmaster = NULL; clusterNode *sender_master = NULL; /* Sender or its master if it is a slave. */
int dirty_slots = 0; /* Sender claimed slots don't match my view? */
/* Here we set curmaster to this node or the node this node if (sender) {
* replicates to if it's a slave. In the for loop we are sender_master = (sender->flags & REDIS_NODE_MASTER) ? sender :
* interested to check if slots are taken away from curmaster. */ sender->slaveof;
if (server.cluster->myself->flags & REDIS_NODE_MASTER) if (sender_master) {
curmaster = server.cluster->myself; dirty_slots = memcmp(sender_master->slots,
else hdr->myslots,sizeof(hdr->myslots)) != 0;
curmaster = server.cluster->myself->slaveof; }
}
for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) { /* 1) We check if the sender (master) is claiming slots that we belive to
if (bitmapTestBit(hdr->myslots,j)) { * serve as a master, or replicate as a slave, but with a configEpoch
/* We rebind the slot to the new node claiming it if: * that is newer: in that case we rebind the slots to the claiming node
* 1) The slot was unassigned. * in our routing table.
* 2) The new node claims it with a greater configEpoch. */ *
if (server.cluster->slots[j] == sender) continue; * Moreover if we are left with 0 slots to serve, we reconfigure as
if (server.cluster->slots[j] == NULL || * a replica of the sender. */
server.cluster->slots[j]->configEpoch < if (sender && sender->flags & REDIS_NODE_MASTER && dirty_slots) {
senderConfigEpoch) int j;
{ clusterNode *curmaster, *newmaster = NULL;
if (server.cluster->slots[j] == curmaster)
newmaster = sender; /* Here we set curmaster to this node or the node this node
clusterDelSlot(j); * replicates to if it's a slave. In the for loop we are
clusterAddSlot(sender,j); * interested to check if slots are taken away from curmaster. */
clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG| if (server.cluster->myself->flags & REDIS_NODE_MASTER)
CLUSTER_TODO_UPDATE_STATE| curmaster = server.cluster->myself;
CLUSTER_TODO_FSYNC_CONFIG); else
} curmaster = server.cluster->myself->slaveof;
} else {
/* This node claims to no longer handling the slot, for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {
* however we don't change our config as this is likely: if (bitmapTestBit(hdr->myslots,j)) {
* 1) Rehashing in progress. /* We rebind the slot to the new node claiming it if:
* 2) Failover. * 1) The slot was unassigned.
* In both cases we'll be informed about who is serving * 2) The new node claims it with a greater configEpoch. */
* the slot eventually. In the meantime it's up to the if (server.cluster->slots[j] == sender) continue;
* original owner to try to redirect our clients to the if (server.cluster->slots[j] == NULL ||
* right node. */ server.cluster->slots[j]->configEpoch <
senderConfigEpoch)
{
if (server.cluster->slots[j] == curmaster)
newmaster = sender;
clusterDelSlot(j);
clusterAddSlot(sender,j);
clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|
CLUSTER_TODO_UPDATE_STATE|
CLUSTER_TODO_FSYNC_CONFIG);
} }
} }
}
/* If at least one slot was reassigned from a node to another node /* If at least one slot was reassigned from a node to another node
* with a greater configEpoch, it is possible that: * with a greater configEpoch, it is possible that:
* 1) We are a master is left without slots. This means that we were * 1) We are a master is left without slots. This means that we were
* failed over and we should turn into a replica of the new * failed over and we should turn into a replica of the new
* master. * master.
* 2) We are a slave and our master is left without slots. We need * 2) We are a slave and our master is left without slots. We need
* to replicate to the new slots owner. */ * to replicate to the new slots owner. */
if (newmaster && curmaster->numslots == 0) { if (newmaster && curmaster->numslots == 0) {
redisLog(REDIS_WARNING,"Configuration change detected. Reconfiguring myself as a replica of %.40s", sender->name); redisLog(REDIS_WARNING,"Configuration change detected. Reconfiguring myself as a replica of %.40s", sender->name);
clusterSetMaster(sender); clusterSetMaster(sender);
clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG| clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|
CLUSTER_TODO_UPDATE_STATE| CLUSTER_TODO_UPDATE_STATE|
CLUSTER_TODO_FSYNC_CONFIG); CLUSTER_TODO_FSYNC_CONFIG);
}
}
/* 2) We also check for the reverse condition, that is, the sender claims
* to serve slots we know are served by a master with a greater
* configEpoch. If this happens we inform the sender.
*
* This is useful because sometimes after a partition heals, a reappearing
* master may be the last one to claim a given set of hash slots, but with
* a configuration that other instances know to be deprecated. Example:
*
* A and B are master and slave for slots 1,2,3.
* A is partitioned away, B gets promoted.
* B is partitioned away, and A returns available.
*
* Usually B would PING A publishing its set of served slots and its
* configEpoch, but because of the partition B can't inform A of the new
* configuration, so other nodes that have an updated table must do it.
* In this way A will stop to act as a master (or can try to failover if
* there are the conditions to win the election). */
if (sender && dirty_slots) {
int j;
for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {
if (bitmapTestBit(hdr->myslots,j)) {
if (server.cluster->slots[j] == sender ||
server.cluster->slots[j] == NULL) continue;
if (server.cluster->slots[j]->configEpoch >
senderConfigEpoch)
{
printf("MASTER or SLAVE have old config\n");
break;
}
} }
} }
} }