int runBug20185(NDBT_Context* ctx, NDBT_Step* step){
int result = NDBT_OK;
int loops = ctx->getNumLoops();
int records = ctx->getNumRecords();
NdbRestarter restarter;
HugoOperations hugoOps(*ctx->getTab());
Ndb* pNdb = GETNDB(step);
int dump[] = { 7090, 20 } ;
if (restarter.dumpStateAllNodes(dump, 2))
return NDBT_FAILED;
NdbSleep_MilliSleep(3000);
if(hugoOps.startTransaction(pNdb) != 0)
return NDBT_FAILED;
if(hugoOps.pkUpdateRecord(pNdb, 1, 1) != 0)
return NDBT_FAILED;
if (hugoOps.execute_NoCommit(pNdb) != 0)
return NDBT_FAILED;
int nodeId;
const int node = hugoOps.getTransaction()->getConnectedNodeId();
do {
nodeId = restarter.getDbNodeId(rand() % restarter.getNumDbNodes());
} while (nodeId == node);
if (restarter.insertErrorInAllNodes(7030))
return NDBT_FAILED;
if (restarter.insertErrorInNode(nodeId, 7031))
return NDBT_FAILED;
NdbSleep_MilliSleep(500);
if (hugoOps.execute_Commit(pNdb) == 0)
return NDBT_FAILED;
NdbSleep_MilliSleep(3000);
restarter.waitClusterStarted();
if (restarter.dumpStateAllNodes(dump, 1))
return NDBT_FAILED;
return NDBT_OK;
}
static int pause_lcp(int error)
{
int nodes = g_restarter.getNumDbNodes();
int filter[] = { 15, NDB_MGM_EVENT_CATEGORY_INFO, 0 };
int fd = ndb_mgm_listen_event(g_restarter.handle, filter);
require(fd >= 0);
require(!g_restarter.insertErrorInAllNodes(error));
int dump[] = { DumpStateOrd::DihStartLcpImmediately };
require(!g_restarter.dumpStateAllNodes(dump, 1));
char *tmp;
char buf[1024];
SocketInputStream in(fd, 1000);
int count = 0;
do {
tmp = in.gets(buf, 1024);
if(tmp)
{
int id;
if(sscanf(tmp, "%*[^:]: LCP: %d ", &id) == 1 && id == error &&
--nodes == 0){
close(fd);
return 0;
}
}
} while(count++ < 30);
close(fd);
return -1;
}
int
resetTransactionTimeout(NDBT_Context* ctx, NDBT_Step* step){
NdbRestarter restarter;
int val[] = { DumpStateOrd::TcSetApplTransactionTimeout, g_org_timeout };
if(restarter.dumpStateAllNodes(val, 2) != 0){
return NDBT_FAILED;
}
return NDBT_OK;
}
int
resetTransactionTimeout(NDBT_Context* ctx, NDBT_Step* step){
NdbRestarter restarter;
// g_org_timeout will be passed as printed int to mgm,
// then converted to Uint32 before sent to tc.
// Check convert Uint32 -> int -> Uint32 is safe
NDB_STATIC_ASSERT(UINT_MAX32 == (Uint32)(int)UINT_MAX32);
int val[] = { DumpStateOrd::TcSetApplTransactionTimeout, (int)g_org_timeout };
if(restarter.dumpStateAllNodes(val, 2) != 0){
return NDBT_FAILED;
}
return NDBT_OK;
}
static int continue_lcp(int error)
{
int filter[] = { 15, NDB_MGM_EVENT_CATEGORY_INFO, 0 };
NDB_SOCKET_TYPE my_fd;
my_socket_invalidate(&my_fd);
#ifdef NDB_WIN
SOCKET fd;
#else
int fd;
#endif
if(error){
fd = ndb_mgm_listen_event(g_restarter.handle, filter);
#ifdef NDB_WIN
my_fd.s= fd;
#else
my_fd.fd= fd;
#endif
require(my_socket_valid(my_fd));
}
int args[] = { DumpStateOrd::LCPContinue };
if(g_restarter.dumpStateAllNodes(args, 1) != 0)
return -1;
if(error){
char *tmp;
char buf[1024];
SocketInputStream in(my_fd, 1000);
int count = 0;
int nodes = g_restarter.getNumDbNodes();
do {
tmp = in.gets(buf, 1024);
if(tmp)
{
int id;
if(sscanf(tmp, "%*[^:]: LCP: %d ", &id) == 1 && id == error &&
--nodes == 0){
my_socket_close(my_fd);
return 0;
}
}
} while(count++ < 30);
my_socket_close(my_fd);
}
return 0;
}
int
setDeadlockTimeout(NDBT_Context* ctx, NDBT_Step* step){
NdbRestarter restarter;
int timeout = ctx->getProperty("TransactionDeadlockTimeout", TIMEOUT);
NdbConfig conf;
if (!conf.getProperty(conf.getMasterNodeId(),
NODE_TYPE_DB,
CFG_DB_TRANSACTION_DEADLOCK_TIMEOUT,
&g_org_deadlock))
return NDBT_FAILED;
g_err << "Setting timeout: " << timeout << endl;
int val[] = { DumpStateOrd::TcSetTransactionTimeout, timeout };
if(restarter.dumpStateAllNodes(val, 2) != 0){
return NDBT_FAILED;
}
return NDBT_OK;
}
int
setTransactionTimeout(NDBT_Context* ctx, NDBT_Step* step){
NdbRestarter restarter;
int timeout = ctx->getProperty("TransactionInactiveTimeout",TIMEOUT);
NdbConfig conf;
if (!conf.getProperty(conf.getMasterNodeId(),
NODE_TYPE_DB,
CFG_DB_TRANSACTION_INACTIVE_TIMEOUT,
&g_org_timeout)){
return NDBT_FAILED;
}
int val[] = { DumpStateOrd::TcSetApplTransactionTimeout, timeout };
if(restarter.dumpStateAllNodes(val, 2) != 0){
return NDBT_FAILED;
}
return NDBT_OK;
}
static int continue_lcp(int error)
{
int filter[] = { 15, NDB_MGM_EVENT_CATEGORY_INFO, 0 };
int fd = -1;
if(error){
fd = ndb_mgm_listen_event(g_restarter.handle, filter);
require(fd >= 0);
}
int args[] = { DumpStateOrd::LCPContinue };
if(g_restarter.dumpStateAllNodes(args, 1) != 0)
return -1;
if(error){
char *tmp;
char buf[1024];
SocketInputStream in(fd, 1000);
int count = 0;
int nodes = g_restarter.getNumDbNodes();
do {
tmp = in.gets(buf, 1024);
if(tmp)
{
int id;
if(sscanf(tmp, "%*[^:]: LCP: %d ", &id) == 1 && id == error &&
--nodes == 0){
close(fd);
return 0;
}
}
} while(count++ < 30);
close(fd);
}
return 0;
}
int
runBug18612SR(NDBT_Context* ctx, NDBT_Step* step){
// Assume two replicas
NdbRestarter restarter;
if (restarter.getNumDbNodes() < 2)
{
ctx->stopTest();
return NDBT_OK;
}
Uint32 cnt = restarter.getNumDbNodes();
for(int loop = 0; loop < ctx->getNumLoops(); loop++)
{
int partition0[256];
int partition1[256];
bzero(partition0, sizeof(partition0));
bzero(partition1, sizeof(partition1));
Bitmask<4> nodesmask;
Uint32 node1 = restarter.getDbNodeId(rand()%cnt);
for (Uint32 i = 0; i<cnt/2; i++)
{
do {
int tmp = restarter.getRandomNodeOtherNodeGroup(node1, rand());
if (tmp == -1)
break;
node1 = tmp;
} while(nodesmask.get(node1));
partition0[i] = node1;
partition1[i] = restarter.getRandomNodeSameNodeGroup(node1, rand());
ndbout_c("nodes %d %d", node1, partition1[i]);
assert(!nodesmask.get(node1));
assert(!nodesmask.get(partition1[i]));
nodesmask.set(node1);
nodesmask.set(partition1[i]);
}
ndbout_c("done");
if (restarter.restartAll(false, true, false))
return NDBT_FAILED;
int dump[255];
dump[0] = 9000;
memcpy(dump + 1, partition0, sizeof(int)*cnt/2);
for (Uint32 i = 0; i<cnt/2; i++)
if (restarter.dumpStateOneNode(partition1[i], dump, 1+cnt/2))
return NDBT_FAILED;
dump[0] = 9000;
memcpy(dump + 1, partition1, sizeof(int)*cnt/2);
for (Uint32 i = 0; i<cnt/2; i++)
if (restarter.dumpStateOneNode(partition0[i], dump, 1+cnt/2))
return NDBT_FAILED;
int val2[] = { DumpStateOrd::CmvmiSetRestartOnErrorInsert, 1 };
if (restarter.dumpStateAllNodes(val2, 2))
return NDBT_FAILED;
if (restarter.insertErrorInAllNodes(932))
return NDBT_FAILED;
if (restarter.startAll())
return NDBT_FAILED;
if (restarter.waitClusterStartPhase(2))
return NDBT_FAILED;
dump[0] = 9001;
for (Uint32 i = 0; i<cnt/2; i++)
if (restarter.dumpStateAllNodes(dump, 2))
return NDBT_FAILED;
if (restarter.waitClusterNoStart(30))
if (restarter.waitNodesNoStart(partition0, cnt/2, 10))
if (restarter.waitNodesNoStart(partition1, cnt/2, 10))
return NDBT_FAILED;
if (restarter.startAll())
return NDBT_FAILED;
if (restarter.waitClusterStarted())
return NDBT_FAILED;
}
return NDBT_OK;
}
int testSlowDihFileWrites(NDBT_Context* ctx, NDBT_Step* step)
{
/* Testcase checks behaviour with slow flushing of DIH table definitions
* This caused problems in the past by exhausting the DIH page pool
* Now there's a concurrent operations limit.
* Check that it behaves with many queued ops, parallel drop/node restarts
*/
/* Run as a 'T1' testcase - do nothing for other tables */
if (strcmp(ctx->getTab()->getName(), "T1") != 0)
return NDBT_OK;
/* 1. Activate slow write error insert
* 2. Trigger LCP
* 3. Wait some time, periodically producing info on
* the internal state
* 4. Perform some parallel action (drop table/node restarts)
* 5. Wait some time, periodically producing info on
* the internal state
* 6. Clear the error insert
* 7. Wait a little longer
* 8. Done.
*/
NdbRestarter restarter;
for (Uint32 scenario = 0; scenario < NUM_SCENARIOS; scenario++)
{
ndbout_c("Inserting error 7235");
restarter.insertErrorInAllNodes(7235);
ndbout_c("Triggering LCP");
int dumpArg = 7099;
restarter.dumpStateAllNodes(&dumpArg, 1);
const Uint32 periodSeconds = 10;
Uint32 waitPeriods = 6;
dumpArg = 7032;
for (Uint32 p=0; p<waitPeriods; p++)
{
if (p == 3)
{
switch ((Scenarios) scenario)
{
case DROP_TABLE:
{
/* Drop one of the early-created tables */
ndbout_c("Requesting DROP TABLE");
ctx->setProperty("DIHWritesRequestType", (Uint32) DROP_TABLE_REQ);
ctx->setProperty("DIHWritesRequest", (Uint32) 1);
break;
}
case RESTART_MASTER:
{
ndbout_c("Requesting Master restart");
ctx->setProperty("DIHWritesRequestType", (Uint32) MASTER_RESTART_REQ);
ctx->setProperty("DIHWritesRequest", (Uint32) 1);
break;
}
case RESTART_SLAVE:
{
ndbout_c("Requesting Slave restart");
ctx->setProperty("DIHWritesRequestType", (Uint32) SLAVE_RESTART_REQ);
ctx->setProperty("DIHWritesRequest", (Uint32) 1);
break;
}
default:
break;
}
}
ndbout_c("Dumping DIH page info to ndbd stdout");
restarter.dumpStateAllNodes(&dumpArg, 1);
NdbSleep_MilliSleep(periodSeconds * 1000);
}
ndbout_c("Clearing error insert...");
restarter.insertErrorInAllNodes(0);
waitPeriods = 2;
for (Uint32 p=0; p<waitPeriods; p++)
{
ndbout_c("Dumping DIH page info to ndbd stdout");
restarter.dumpStateAllNodes(&dumpArg, 1);
NdbSleep_MilliSleep(periodSeconds * 1000);
}
ndbout_c("Waiting for worker to finish task...");
ctx->getPropertyWait("DIHWritesRequest", 2);
if (ctx->isTestStopped())
return NDBT_OK;
ndbout_c("Done.");
}
/* Finish up */
ctx->stopTest();
return NDBT_OK;
}
int
NdbBackup::Fail(NdbRestarter& _restarter, int *Fail_codes, const int sz, bool onMaster){
CHECK(_restarter.waitClusterStarted() == 0,
"waitClusterStarted failed");
int nNodes = _restarter.getNumDbNodes();
myRandom48Init(NdbTick_CurrentMillisecond());
for(int i = 0; i<sz; i++){
int error = Fail_codes[i];
unsigned int backupId;
const int masterNodeId = _restarter.getMasterNodeId();
CHECK(masterNodeId > 0, "getMasterNodeId failed");
int nodeId;
nodeId = masterNodeId;
if (!onMaster) {
int randomId;
while (nodeId == masterNodeId) {
randomId = myRandom48(nNodes);
nodeId = _restarter.getDbNodeId(randomId);
}
}
g_err << "NdbBackup::Fail node = " << nodeId
<< " error code = " << error << " masterNodeId = "
<< masterNodeId << endl;
CHECK(_restarter.insertErrorInNode(nodeId, error) == 0,
"failed to set error insert");
g_info << "error inserted" << endl;
g_info << "waiting some before starting backup" << endl;
g_info << "starting backup" << endl;
int r = start(backupId);
g_info << "r = " << r
<< " (which should fail) started with id = " << backupId << endl;
if (r == 0) {
g_err << "Backup should have failed on error_insertion " << error << endl
<< "Master = " << masterNodeId << "Node = " << nodeId << endl;
return NDBT_FAILED;
}
CHECK(_restarter.waitClusterStarted() == 0,
"waitClusterStarted failed");
CHECK(_restarter.insertErrorInNode(nodeId, 10099) == 0,
"failed to set error insert");
NdbSleep_SecSleep(5);
int val2[] = { 24, 2424 };
CHECK(_restarter.dumpStateAllNodes(val2, 2) == 0,
"failed to check backup resources RestartOnErrorInsert");
}
return NDBT_OK;
}
int
NdbBackup::NF(NdbRestarter& _restarter, int *NFDuringBackup_codes, const int sz, bool onMaster){
int nNodes = _restarter.getNumDbNodes();
{
if(nNodes == 1)
return NDBT_OK;
int nodeId = _restarter.getMasterNodeId();
CHECK(_restarter.restartOneDbNode(nodeId, false, true, true) == 0,
"Could not restart node "<< nodeId);
CHECK(_restarter.waitNodesNoStart(&nodeId, 1) == 0,
"waitNodesNoStart failed");
CHECK(_restarter.startNodes(&nodeId, 1) == 0,
"failed to start node");
}
CHECK(_restarter.waitClusterStarted() == 0,
"waitClusterStarted failed");
myRandom48Init(NdbTick_CurrentMillisecond());
for(int i = 0; i<sz; i++){
int error = NFDuringBackup_codes[i];
unsigned int backupId;
const int masterNodeId = _restarter.getMasterNodeId();
CHECK(masterNodeId > 0, "getMasterNodeId failed");
int nodeId;
nodeId = masterNodeId;
if (!onMaster) {
int randomId;
while (nodeId == masterNodeId) {
randomId = myRandom48(nNodes);
nodeId = _restarter.getDbNodeId(randomId);
}
}
g_err << "NdbBackup::NF node = " << nodeId
<< " error code = " << error << " masterNodeId = "
<< masterNodeId << endl;
int val[] = { DumpStateOrd::CmvmiSetRestartOnErrorInsert, 1 };
CHECK(_restarter.dumpStateOneNode(nodeId, val, 2) == 0,
"failed to set RestartOnErrorInsert");
CHECK(_restarter.insertErrorInNode(nodeId, error) == 0,
"failed to set error insert");
g_info << "error inserted" << endl;
NdbSleep_SecSleep(1);
g_info << "starting backup" << endl;
int r = start(backupId);
g_info << "r = " << r
<< " (which should fail) started with id = " << backupId << endl;
if (r == 0) {
g_err << "Backup should have failed on error_insertion " << error << endl
<< "Master = " << masterNodeId << "Node = " << nodeId << endl;
return NDBT_FAILED;
}
CHECK(_restarter.waitNodesNoStart(&nodeId, 1) == 0,
"waitNodesNoStart failed");
g_info << "number of nodes running " << _restarter.getNumDbNodes() << endl;
if (_restarter.getNumDbNodes() != nNodes) {
g_err << "Failure: cluster not up" << endl;
return NDBT_FAILED;
}
g_info << "starting new backup" << endl;
CHECK(start(backupId) == 0,
"failed to start backup");
g_info << "(which should succeed) started with id = " << backupId << endl;
g_info << "starting node" << endl;
CHECK(_restarter.startNodes(&nodeId, 1) == 0,
"failed to start node");
CHECK(_restarter.waitClusterStarted() == 0,
"waitClusterStarted failed");
g_info << "node started" << endl;
int val2[] = { 24, 2424 };
CHECK(_restarter.dumpStateAllNodes(val2, 2) == 0,
"failed to check backup resources RestartOnErrorInsert");
CHECK(_restarter.insertErrorInNode(nodeId, 10099) == 0,
"failed to set error insert");
NdbSleep_SecSleep(1);
}
return NDBT_OK;
}