JobStatus FullSnapJob::onSyncData(bool bto, int res)
{
if(bto || res != 0){
return onFailure(bto,res);
}else{
for(std::vector<Meta>::const_iterator it = infos.begin(); it != infos.end(); ++it){
const Meta& m = *it;
if((m.serverId&0xFFF) == cli->nv->getUniqNodeId())
{
if(((m.serverId>>16) != 0))
{
LogTrace(Error, "NodeId conflict, please have a check,m.serverId = %lld,mynodeId = %d",m.serverId, cli->nv->getUniqNodeId());
complete = true;
cli->snapGameover = true;
return JOBSTOP;
}
else
{
LogTrace(Warn, "meta about my service, serverId = %lld, name = %s, groupId = %d", m.serverId, m.name.c_str(), m.groupId);
}
}
log(Debug, "set metaId:%lld ts:%lld", m.serverId, m.timestamp);
int ret = cli->mDao->set(m, false);
if(ret < 0 ){
LogTrace(Error, "meta set error, errno = %d", myerrno);
if(myerrno == ETIMESTAMPLT)
{
LogTrace(Error, "this is empty node, meta should not be timeout");
complete = true;
cli->snapGameover = true;
return JOBSTOP;
}
ASYNC_CALL(System, sleep, &FullSnapJob::onWakeup, 1000 * 10);
return SYSDEFAULT;
}
if(m.isDel == 0)
{
syncUndelMetaCount++;
}
else
{
syncDelMetaCount++;
}
lastServerId = it->serverId;
}
size_t sz = infos.size();
LogTrace(Info, "sync meta data, size = %d,lastServerId = %lld ", infos.size(), lastServerId);
infos.clear();
if(sz < 1000){
step = SyncLastLsn_C;
ASYNC_CALL(SnapService, getLsns, &FullSnapJob::onSyncLsn, logs, sessionId);
}else{
ASYNC_CALL(SnapService, getMetas, &FullSnapJob::onSyncData, infos, sessionId, lastServerId, 1000);
}
return SYSDEFAULT;
}
JobStatus FullSnapJob::onSyncLog(bool bto, int res)
{
if(bto || res != 0){
return onFailure(bto,res);
}else{
size_t sz = logs.size();
bool applyErr = false;
size_t i = 0;
for(; i < sz; ++i){
const DLog& dlog = logs[i];
int ret = cli->mDao->apply(dlog,false);
if(ret<0)
{
LogTrace(Error, "sync log apply error, errno = %d", myerrno);
applyErr = true;
break;
}
lastLsnId = dlog.lsn;
}
syncLogCount += i; // count the stored DLog;
LogTrace(Info, "sync log, sz = %d, total = %d", sz, syncLogCount);
logs.clear();
if(applyErr)
{
ASYNC_CALL(System, sleep, &FullSnapJob::onWakeup, 1000 * 10);
return SYSDEFAULT;
}
if(sz < 1000)
{
lsnVecIndex++;
lastLsnId = 0;
if(lsnVecIndex < cli->lastLsn.size()) // sync Dlog of next nodeid
{
ASYNC_CALL(SnapService, getDLogs, &FullSnapJob::onSyncLog, logs, sessionId,cli->lastLsn[lsnVecIndex].node,lastLsnId,1000);
}
else // sync finished, stop
{
ASYNC_CALL(SnapService, snapStop, &FullSnapJob::onStop, stop, sessionId);
return SYSDEFAULT;
}
}
else
{
ASYNC_CALL(SnapService, getDLogs, &FullSnapJob::onSyncLog, logs, sessionId,cli->lastLsn[lsnVecIndex].node,lastLsnId,1000);
}
return SYSDEFAULT;
}
}
JobStatus FullSnapJob::onSyncLsn(bool bto, int res)
{
if(bto || res != 0){
return onFailure(bto, res);
}else{
std::vector<Lsn> tmpLsns;
for(std::vector<DLog>::const_iterator it = logs.begin(); it != logs.end(); ++it){
Lsn tmp;
tmp.node = it->node;
tmp.lsn = it->lsn;
tmpLsns.push_back(tmp);
LogTrace(Info, "node = %d, lsn = %lld",tmp.node, tmp.lsn);
}
logs.clear();
cli->lastLsn.swap(tmpLsns); // update session sync's lastlsns
if(cli->lastLsn.empty())
{
LogTrace(Error,"last lsn is empty");
complete = true;
cli->snapGameover = true;
return JOBSTOP;
}
lsnVecIndex = 0;
lastLsnId = 0;
step = SyncDLog_C;
ASYNC_CALL(SnapService, getDLogs, &FullSnapJob::onSyncLog, logs, sessionId,cli->lastLsn[lsnVecIndex].node,lastLsnId,1000);
return SYSDEFAULT;
}
}
static void colo_proxy_wait_for_checkpoint(libxl__egc *egc,
libxl__colo_save_state *css)
{
libxl__domain_save_state *dss = CONTAINER_OF(css, *dss, css);
ASYNC_CALL(egc, dss->cds.ao, &css->child, css,
colo_proxy_async_wait_for_checkpoint,
colo_proxy_async_call_done);
}
JobStatus FullSnapJob::doReconnect()
{
LogTrace(Info,"Method Entry");
failCount = 0;
syncLogCount = 0;
syncUndelMetaCount = 0;
syncDelMetaCount = 0;
cli->lastLsn.clear();
int port = 0;
curIp.clear();
if(cli->cfg->isLocalDaemon())
{
if(ipIndex < cli->ipListFromDns.size())
{
curIp = cli->ipListFromDns[ipIndex];
port = cli->cfg->cSnapPort;
}
}
else // daemon
{
if(ipIndex < cli->ipListFromXml.size()){ // after DNS,select a new ip;
curIp = cli->ipListFromXml[ipIndex].ip;
port = cli->ipListFromXml[ipIndex].port;
targetNode = cli->ipListFromXml[ipIndex].gid;
}
}
ipIndex++;
if(curIp.empty())
{
LogTrace(Error,"failure from all daemons, please have a check");
cli->snapGameover = true;
return JOBSTOP;
}
LogTrace(Info,"curIP = %s, isLocalDaemon = %d", curIp.c_str(), cli->cfg->isLocalDaemon());
IAsyncCenter *center = cli->center;
TRowTransportFactory *to = new TRowTransportFactory();
to->init(center->getClientConnFactory(), ctxMgr->getHandler(), curIp, port);
to->setMainThreadQueue(center->getMainQueue());
tf.reset(to);
SETTRANSPORT(SnapService, tf);
step = SyncNotStart_C;
ASYNC_CALL(SnapService, snapStart, &FullSnapJob::onStarted, sessionId, cli->nv->getGid(), targetNode);
return SYSDEFAULT;
}
JobStatus FullSnapJob::onWakeup(bool bto, int res)
{
if(step == SyncMetaData_C)
{
ASYNC_CALL(SnapService, getMetas, &FullSnapJob::onSyncData, infos, sessionId, lastServerId, 1000);
}
else if(step == SyncDLog_C)
{
ASYNC_CALL(SnapService, getDLogs, &FullSnapJob::onSyncLog, logs, sessionId,cli->lastLsn[lsnVecIndex].node,lastLsnId,1000);
}
else if(step == SyncLastLsn_C)
{
ASYNC_CALL(SnapService, getLsns, &FullSnapJob::onSyncLsn, logs, sessionId);
}
else
{
LogTrace(Error,"invalid step = %d", step);
complete = true;
cli->snapGameover = true;
return JOBSTOP;
}
return SYSDEFAULT;
}
JobStatus FullSnapJob::onStarted(bool bto, int res)
{
if(bto || res != 0){
LogTrace(Warn, "time out or err, bto = %d, res = %d, mine = %d,targetNode = %d", bto, res, cli->nv->getGid(), targetNode);
return doReconnect();
}else{
if(startCount >0)
{
LogTrace(Warn, "prev snap failure, truncate the DB");
cli->mDao->truncate();
}
lastServerId = 0;
step = SyncMetaData_C;
startCount++;
ASYNC_CALL(SnapService, getMetas, &FullSnapJob::onSyncData, infos, sessionId, lastServerId, 1000);
return SYSDEFAULT;
}
}
