Lines Matching defs:node
354 * sent the message to the wrong node. Leave fixing that for
804 * If a MN diskset and snarfing and this node is not the
808 * Master node should have already handled this case.
833 * Allocate pernode bitmap for this node. All other nodes' maps will
901 * MD_RI_BLOCK_OWNER is set by the ownership change of a multi-node mirror.
1031 * when the resync completes. This happens because only one node is
1089 /* if the node hasn't yet joined, it's Ok. */
1161 * did not pass this information here, the only time every node gets
1272 * Check to see if another node has completed this
1307 * Check to see if another node has completed
1347 * We aren't marked as needing a resync so for multi-node
1349 * metadevice state. This is a problem when a new node joins
1353 * have the metadevices marked as 'Okay' but the joining node
1392 * resync. If not, another node must have completed it
1402 * resync that has already been partially done on another node.
1437 * resync. If not, another node must have completed it
1448 * blocked in resync_read_blk_range while another node
1451 * node and update rr to the next RR to be done.
1532 * resync on a single node having rebooted with a different value of
1533 * md_resync_bufsz or if we are running in a multi-node with nodes having
1562 * If un_rs_startbl is set, another node must have already started
1650 * either because this node has been rebooted with a different value
1651 * for md_resync_bufsz or because another node with a different value
1703 * resync. If not, another node must have completed it
1715 * another node, with a different resync buffer size and it may have
1753 * component resync. If not, another node must
1758 * node continued with the resync and
1771 * have lost ownership to another node, with a
1802 * another node has progressed the resync.
1814 * another node, with a different resync buffer size and it may
1823 * resync_read_blk_range while another node continued
1832 * If we are picking up from another node, we retry the last
1973 * resync. If not, another node must have completed it
2028 * as another node has completed the submirror resync.
2037 * in resync_read_blk_range while another node
2145 * Now check if another node has continued with the
2212 * Commit mirror unit if we're the Master node in a multi-node
2257 * submirror). Must handle thread suspension and kill to allow multi-node
2265 * master node sends a CHOOSE_OWNER message to the handler on the master. This
2267 * selected node to become the owner.
2268 * If this node is not the owner it sets itself to block in resync_kill_pending
2271 * function only one node will continue past resync_kill_pending().
2282 * block size to limit the amount of inter node traffic. The RESYNC_NEXT
2285 * that no node can write to a block that is being resynced. For all MN
2286 * mirrors we also block the whole resync region on the resync owner node so
2296 * owner and the node that is writing to the mirror, we move the resync to the
2301 * has this must be because we have lost ownership to another node(s) for a
2310 * non resync-owner node without a change of ownership. This is the case when
2409 * initiate an ownership change to this node and then
2441 * the resync count will be used to choose the next node to
2550 * If another node has moved the resync on, we must
2586 * node must have completed it and started a submirror
2588 * multi-node sets we move onto the submirror resync.
2613 * node must have completed it and started a different
2650 * For MN Set, send a RESYNC_FINISH if this node completed the resync.
2659 * message if we're part of a multi-node set.
3075 * flag is set on a node when it is being added to a diskset
3076 * so that the resync threads are started on the newly added node.
3150 * message to the owning node so that the
3194 * about another node twice won't hurt anything.
3632 * Set the pernode_dirty_bm[node] entries and un_dirty_bm[] if appropriate.
3633 * For the owning node (DRL/mirror owner) update the on-disk RR if needed.
3634 * Called on every clean->dirty transition for the originating writer node.
3636 * the owning node that has to process it.
3689 * Clear the pernode_dirty_bm[node] entries which are passed in the bitmap
3702 md_mn_nodeid_t node = iocp->rr_nodeid - 1;
3711 rw_enter(&un->un_pernode_dirty_mx[node], RW_READER);
3725 rw_exit(&un->un_pernode_dirty_mx[node]);
3782 * Clear the pernode_dirty_bm[node] entries which are passed in the bitmap
3798 md_mn_nodeid_t node = iocp->rr_nodeid - 1;
3838 rw_enter(&un->un_pernode_dirty_mx[node], RW_WRITER);
3839 if (un->un_pernode_dirty_bm[node] == NULL) {
3840 un->un_pernode_dirty_bm[node] = (uchar_t *)kmem_zalloc(
3843 rw_exit(&un->un_pernode_dirty_mx[node]);
3847 * the issuing node _and_ we aren't in the process of marking/clearing
3849 * If we're the owning node and _not_ the issuing node, we should also
3867 * we must schedule a blocking call to update the DRL on this node.