wbtiger
/
foundationdb
mirror of https://github.com/apple/foundationdb.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

fixed bug in maintaining kill set size

This commit is contained in:
Jon Fu 2019-11-04 13:49:32 -08:00
parent b326d26a47
commit f7b3686fc7
1 changed files with 21 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
fdbserver/workloads/RemoveServersSafely.actor.cpp
View File

@ -452,18 +452,35 @@ struct RemoveServersSafelyWorkload : TestWorkload {
// Swap coordinator with one server in the kill set to ensure the number of processes to kill does not increase.
// This is needed only if a new coordinator is added to the toKill set in this function and safety check passes
if (markExcludeAsFailed && coordExcl.isValid()) {
// Situation where the entirety of original kill set is selected and extra coordinator is added
// Shrink down failed vector to maintain size guarantees
if (toKillMarkFailedArray.size() > toKillArray.size()) {
auto removeServer = toKillMarkFailedArray.begin();
TraceEvent("RemoveAndKill", functionId)
.detail("Step", "ShrinkFailedKillSet")
.detail("Removing", removeServer->toString());
toKillMarkFailedArray.erase(removeServer);
}
auto removeServer = toKill.begin();
TraceEvent("RemoveAndKill", functionId)
.detail("Step", "ReplaceKillSet")
.detail("Step", "ReplaceNonFailedKillSet")
.detail("Removing", removeServer->toString())
.detail("Adding", coordExcl.toString());
toKill.erase(removeServer);
toKill.insert(coordExcl);
toKillArray.erase(std::remove(toKillArray.begin(), toKillArray.end(), *removeServer), toKillArray.end());
toKillArray.push_back(coordExcl);
toKill.erase(removeServer);
toKill.insert(coordExcl);
}
killProcArray = self->getProcesses(toKill);
TraceEvent("RemoveAndKill", functionId).detail("Step", "Activate Server Exclusion").detail("KillAddrs", toKill.size()).detail("KillProcs", killProcArray.size()).detail("MissingProcs", toKill.size()!=killProcArray.size()).detail("ToKill", describe(toKill)).detail("Addresses", describe(toKillArray)).detail("ClusterAvailable", g_simulator.isAvailable());
TraceEvent("RemoveAndKill", functionId)
.detail("Step", "Activate Server Exclusion")
.detail("KillAddrs", toKill.size())
.detail("KillProcs", killProcArray.size())
.detail("MissingProcs", toKill.size() != killProcArray.size())
.detail("ToKill", describe(toKill))
.detail("Addresses", describe(toKillArray))
.detail("FailedAddresses", describe(toKillMarkFailedArray))
.detail("ClusterAvailable", g_simulator.isAvailable());
if (markExcludeAsFailed) {
wait( excludeServers( cx, toKillMarkFailedArray, true ) );
}