Debian Codegraph @

mysql-test/suite/ndb/r/ndb_TCtakeover_stall.result

raw
1 A client holding a row lock can cause node restart
2 (CopyFrag) to stall until the row lock is released.
3
4 Check that a stalled CopyFrag does not lead to
5 TC takeover stall by :
6 - Restarting a node
7 - Stalling CopyFrag due to a rowlock
8 - Causing starting node failure
9 - Checking that TC takeover completes
10 (As transactions can be committed in
11 different epochs)
12 create table t1(id int primary key, val int) engine=ndb;
13 select count(*) from t1;
14 count(*)
15 10
16 begin;
17 select * from t1 where id = 1 for update;
18 id val
19 1 1
20 select tc_node_id from ndbinfo.cluster_operations;
21 tc_node_id
22 1
23 select ndbinfo.cluster_transactions.state,
24 ndbinfo.cluster_operations.operation_type, ndbinfo.cluster_operations.state,
25 ndbinfo.cluster_locks.mode
26 from ndbinfo.cluster_transactions, ndbinfo.cluster_operations,
27 ndbinfo.cluster_locks
28 where
29 ndbinfo.cluster_transactions.transid = ndbinfo.cluster_locks.transid and
30 ndbinfo.cluster_locks.transid = ndbinfo.cluster_operations.transid;
31 state operation_type state mode
32 Started READ-EX Prepared X
33 Starting the nowait node
34 Eventually it will be waiting for the lock above
35 Kill the starting node
36 This will result in TC takeover
37 Sleeping 10 s for node failure to be detected
38 and handling to start [and finish]
39 Test succeeded, no GCP stall
40 Release lock to allow data node start to succeed
41 rollback;
42 Wait for all data nodes to start
43 Started ok
44 drop table t1;