Rongsong Shen
65ddafeb66
Append-only UInt64 epoch-millis field for the responder snapshot timestamp (Req 11.1, wire-format setup). Hand-authored xce_ctrlstatus_t typemap gains a matching responder_time_ms member; codec round-trips it. Controller suite 1/1 OK. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
894 lines
32 KiB
Cap'n Proto
894 lines
32 KiB
Cap'n Proto
@0xdefcdf6377db2e8e;
|
||
|
||
using C = import "/c.capnp";
|
||
|
||
$C.fieldgetset;
|
||
$C.codecgen;
|
||
$C.extraheader("#include \"clusterconfig.h\"");
|
||
$C.extraheader("#include \"typemap.h\"");
|
||
|
||
enum OPCode {
|
||
nop @0;
|
||
reply @1;
|
||
hello @2;
|
||
join @3;
|
||
shutdown @4;
|
||
get @5;
|
||
push @6;
|
||
update @7;
|
||
jobdispatch @8;
|
||
startjob @9;
|
||
notify @10;
|
||
jobsub @11;
|
||
jobstatus @12;
|
||
jobrusage @13;
|
||
hostcontrol @14;
|
||
jobcontrol @15;
|
||
newschedjobs @16;
|
||
clusterconf @17;
|
||
queuecontrol @18;
|
||
runtasks @19;
|
||
setpreason @20;
|
||
clrpreason @21;
|
||
replEntry @22;
|
||
replAck @23;
|
||
redirect @24;
|
||
heartbeat @25;
|
||
steppingDown @26;
|
||
replSnapshot @27;
|
||
replCatchUpDone @28;
|
||
replConfig @29;
|
||
replConfigPush @30;
|
||
upgradeConfig @31;
|
||
requestVote @32;
|
||
probeConfigTag @33;
|
||
walAck @34;
|
||
upgradeConfigAck @35;
|
||
snapshotTransfer @36;
|
||
walCatchupEntry @37;
|
||
preVoteRequest @38;
|
||
preVoteReply @39;
|
||
ctrlstatus @40;
|
||
}
|
||
|
||
struct PeerInfo $C.mapname("xce_peerinfo_t") {
|
||
host @0: Text $C.mapname("name");
|
||
key @1: Text;
|
||
}
|
||
|
||
struct HostBasedReason $C.mapname("xce_host_reason_t") {
|
||
host @0: Text;
|
||
reason @1: UInt32;
|
||
}
|
||
|
||
struct ResourceBasedReason $C.mapname("xce_res_reason_t") {
|
||
resource @0: Text;
|
||
host @1: Text;
|
||
reason @2: UInt32;
|
||
}
|
||
|
||
struct JobPendReason $C.mapname("xce_job_preason_t") {
|
||
mainReason @0: UInt32 $C.mapname("main_reason");
|
||
hostReasons @1: List(HostBasedReason) $C.mapname("host_based_reasons") $C.maplistcount("nhost_reasons");
|
||
resReasons @2: List(ResourceBasedReason) $C.mapname("res_based_reasons") $C.maplistcount("nres_reasons");
|
||
}
|
||
|
||
struct StrList $C.mapname("xce_strlist_t") {
|
||
data @0: List(Text) $C.mapname("data") $C.maplistcount("count");
|
||
}
|
||
|
||
struct TimeWindowTime $C.mapname("xce_timewindow_time_t") {
|
||
day @0: UInt32 $C.mapname("day");
|
||
hour @1: UInt32 $C.mapname("hour");
|
||
min @2: UInt32 $C.mapname("min");
|
||
}
|
||
|
||
struct TimeWindowSeg $C.mapname("xce_timewindow_seg_config_t") {
|
||
timezone @0: Text $C.mapname("tz");
|
||
open @1: TimeWindowTime $C.mapname("open");
|
||
close @2: TimeWindowTime $C.mapname("close");
|
||
}
|
||
|
||
struct TimeWindow $C.mapname("xce_timewindow_config_t") {
|
||
segs @0: List(TimeWindowSeg) $C.mapname("segs") $C.maplistcount("nsegs");
|
||
}
|
||
|
||
struct GroupConfig $C.mapname("xce_group_config_t") {
|
||
name @0: Text $C.mapname("name");
|
||
members @1: List(Text) $C.mapname("members.data") $C.maplistcount("members.count");
|
||
exmembers @2: List(Text) $C.mapname("ex_members.data") $C.maplistcount("ex_members.count");
|
||
}
|
||
|
||
# This data structure will not be transfer in wire
|
||
# directly, it will be assembled by multiple Data packet
|
||
struct Dataset $C.mapname("dataset_t") {
|
||
table @0: UInt32;
|
||
nrows @1: UInt32;
|
||
rows @2: Void; # do NOT change this, it's place holder
|
||
}
|
||
|
||
struct ResKey $C.mapname("xce_reskey_t") {
|
||
resname @0: Text;
|
||
hostname @1: Text;
|
||
}
|
||
|
||
struct MatchCondition $C.mapname("xce_match_t") {
|
||
condition :union $C.mapuniontag("type") {
|
||
matchall @0: Void;
|
||
matchjid @1: UInt64 $C.mapname("jobid");
|
||
matchwname @2: Text $C.mapname("wname");
|
||
matchreskey @3: ResKey $C.mapname("reskey");
|
||
}
|
||
}
|
||
|
||
struct HostAttrs $C.mapname("hostattrs_t") {
|
||
type @0: Text $C.mapname("hosttype");
|
||
model @1: Text $C.mapname("hostmodel");
|
||
cpufactor @2: Float64;
|
||
maxcpus @3: Int32;
|
||
maxmem @4: Float64;
|
||
maxswap @5: Float64;
|
||
maxtmp @6: Float64;
|
||
ndisks @7: Int32;
|
||
}
|
||
|
||
struct Worker $C.mapname("xce_worker_t"){
|
||
name @0: Text;
|
||
attrs @1: HostAttrs;
|
||
status @2: UInt32;
|
||
dispatchWindow @3: TimeWindow $C.mapname("dispatch_window_config");
|
||
thresholds @4: List(Threshold) $C.mapname("thresholds") $C.maplistcount("nthresholds");
|
||
resvals @5: List(NameValue) $C.mapname("resvals_to_sched") $C.maplistcount("nresvals_to_sched");
|
||
deadSince @6: UInt64 = 0 $C.mapname("dead_since"); # epoch seconds; set when worker becomes UNAVAILABLE
|
||
}
|
||
|
||
enum HostRole {
|
||
client @0;
|
||
worker @1;
|
||
scheduler @2;
|
||
controller @3;
|
||
apiserver @4;
|
||
}
|
||
|
||
struct JoinData $C.mapname("xce_joindata_t") {
|
||
role @0: HostRole;
|
||
data :union {
|
||
nodata @1: Void;
|
||
attrs @2: HostAttrs;
|
||
}
|
||
}
|
||
|
||
struct Value $C.mapname("xce_val_t"){
|
||
u :union $C.mapuniontag("which") {
|
||
ivalue32 @0: Int32;
|
||
uvalue32 @1: UInt32;
|
||
ivalue64 @2: Int64;
|
||
uvalue64 @3: UInt64;
|
||
svalue @4: Text;
|
||
fvalue @5: Float64;
|
||
bvalue @6: Bool;
|
||
}
|
||
}
|
||
|
||
struct LimitValue $C.mapname("xce_limit_val_t") {
|
||
resname @0: Text $C.mapname("resname");
|
||
percentage @1: Bool $C.mapname("percent");
|
||
value @2: Float64 $C.mapname("val");
|
||
}
|
||
|
||
struct ValueWithIdx $C.mapname("xce_val_with_idx_t") {
|
||
idx @0: UInt32;
|
||
val @1: Value $C.mapname("val");
|
||
}
|
||
|
||
struct NameValue $C.mapname("xce_name_val_t") {
|
||
name @0: Text $C.mapname("name");
|
||
val @1: Value $C.mapname("val");
|
||
}
|
||
|
||
struct ConfigVar $C.mapname("xce_config_t") {
|
||
name @0: Text;
|
||
value @1: Text;
|
||
}
|
||
|
||
struct ResourceDistribution $C.mapname("res_distribution_t") {
|
||
value @0: Value;
|
||
hosts @1: List(Text) $C.maplistcount("nhosts");
|
||
}
|
||
|
||
struct Resource $C.mapname("xce_resource_t"){
|
||
name @0: Text;
|
||
desc @1: Text;
|
||
idx @2: Int32;
|
||
valtype @3: Int32;
|
||
ordtype @4: Int32;
|
||
flags @5: Int32;
|
||
interval @6: Int32;
|
||
lastReport @7: UInt64 $C.mapname("last_report");
|
||
# distributions @8: List(ResourceDistribution) $C.maplistcount("ndistributions");
|
||
# hostBased @9: Value $C.mapname("hbased_defined");
|
||
}
|
||
|
||
struct Threshold $C.mapname("xce_res_threshold_t"){
|
||
name @0: Text;
|
||
sched @1: Float64;
|
||
stop @2: Float64;
|
||
}
|
||
|
||
struct QueueConf $C.mapname("xce_queue_t") {
|
||
name @0: Text $C.mapname("queue");
|
||
attrs @1: UInt32;
|
||
status @2: UInt32;
|
||
priority @3: Int32;
|
||
acceptIntervals @4: Int32 $C.mapname("accept_interval");
|
||
schedDelay @5: Int32 $C.mapname("sched_delay");
|
||
resreq @6: Text;
|
||
users @7: GroupConfig $C.mapname("users");
|
||
hosts @8: GroupConfig $C.mapname("hosts");
|
||
thresholds @9: List(Threshold) $C.maplistcount("nthresholds");
|
||
preExec @10: Text $C.mapname("pre_exec");
|
||
postExec @11: Text $C.mapname("post_exec");
|
||
runWindow @12: TimeWindow $C.mapname("runwindow_config");
|
||
dispatchWindow @13: TimeWindow $C.mapname("dispatchwindow_config");
|
||
}
|
||
|
||
struct Conf4Sched $C.mapname("xce_conf4sched_t"){
|
||
clustername @0: Text;
|
||
configvars @1: List(ConfigVar) $C.mapname("confvars") $C.maplistcount("n_confvars");
|
||
resources @2: List(Resource) $C.maplistcount("n_resources");
|
||
queues @3: List(QueueConf) $C.maplistcount("n_queues");
|
||
ugroups @4: List(GroupConfig) $C.maplistcount("n_ugroups");
|
||
hgroups @5: List(GroupConfig) $C.maplistcount("n_hgroups");
|
||
resvals @6: List(NameValue) $C.mapname("resvals") $C.maplistcount("nresvals");
|
||
hproviders @7: List(ScalerConfig) $C.mapname("hproviders") $C.maplistcount("nhproviders");
|
||
limits @8: List(LimitConfig) $C.mapname("limits") $C.maplistcount("nlimits");
|
||
}
|
||
|
||
struct Conf4Worker $C.mapname("xce_conf4worker_t") {
|
||
clustername @0: Text;
|
||
configvars @1: List(ConfigVar) $C.mapname("confvars") $C.maplistcount("n_confvars");
|
||
}
|
||
|
||
struct ConfigData $C.mapname("xce_configdata_t") {
|
||
data :union $C.mapuniontag("data_which") {
|
||
schedconfig @0: Conf4Sched $C.mapname("conf4sched");
|
||
workerconfig @1: Conf4Worker $C.mapname("conf4worker");
|
||
clusterconfig @2: ClusterConfig $C.mapname("clusterconfig");
|
||
}
|
||
}
|
||
|
||
struct JobPIDInfo $C.mapname("job_pidinfo_t"){
|
||
jobId @0: UInt64 $C.mapname("jobid");
|
||
pids @1: List(UInt32) $C.mapname("pids") $C.maplistcount("npids");
|
||
pgids @2: List(UInt32) $C.mapname("pgids") $C.maplistcount("npgids");
|
||
exdata @3: Text $C.mapname("exdata");
|
||
}
|
||
|
||
struct JobFail $C.mapname("job_failinfo_t"){
|
||
jobId @0: UInt64 $C.mapname("jobid");
|
||
code @1: Int32 $C.mapname("code");
|
||
desc @2: Text $C.mapname("detail");
|
||
stage @3: UInt32 $C.mapname("stage");
|
||
}
|
||
|
||
struct JobStartInfo $C.mapname("job_startinfo_t"){
|
||
exechome @0: Text;
|
||
execcwd @1: Text;
|
||
execuser @2: Text;
|
||
cmdline @3: Text;
|
||
exdata @4: Text;
|
||
launchtype @5: Text $C.mapname("launch_type");
|
||
qPreExec @6: Text $C.mapname("q_pre_exec");
|
||
qPostExec @7: Text $C.mapname("q_post_exec");
|
||
preExec @8: Text $C.mapname("job_pre_exec");
|
||
postExec @9: Text $C.mapname("job_post_exec");
|
||
}
|
||
|
||
struct Job $C.mapname("xce_job_t"){
|
||
jobid @0: UInt64;
|
||
submitData @1: JobSubmitData $C.mapname("submit");
|
||
subtime @2: UInt64 $C.mapname("sub_time");
|
||
queuetime @3: UInt64 $C.mapname("queue_time");
|
||
starttime @4: UInt64 $C.mapname("start_time");
|
||
finishtime @5: UInt64 $C.mapname("finish_time");
|
||
state @6: UInt32;
|
||
allocs @7: JobAlloc $C.mapname("allocs");
|
||
runtime @8: UInt32 $C.mapname("run_time");
|
||
pidinfo @9: JobPIDInfo;
|
||
pendreason @10: JobPendReason $C.mapname("pend_reasons");
|
||
transientSince @11: UInt64 = 0 $C.mapname("transient_since"); # epoch seconds; set when job enters JOB_STAT_TRANSIENT
|
||
}
|
||
|
||
struct SchedJob $C.mapname("xce_job_t"){
|
||
jobid @0: UInt64;
|
||
submitData @1: JobSubmitData $C.mapname("submit");
|
||
subtime @2: UInt64 $C.mapname("sub_time");
|
||
queuetime @3: UInt64 $C.mapname("queue_time");
|
||
starttime @4: UInt64 $C.mapname("start_time");
|
||
finishtime @5: UInt64 $C.mapname("finish_time");
|
||
state @6: UInt32;
|
||
allocs @7: JobAlloc $C.mapname("allocs");
|
||
runtime @8: UInt32 $C.mapname("run_time");
|
||
}
|
||
|
||
struct SchedJobRusages $C.mapname("xce_job_rusages_t") {
|
||
jobid @0: UInt64 $C.mapname("jobid");
|
||
usages @1: List(ResUsage) $C.mapname("usages") $C.maplistcount("count");
|
||
}
|
||
|
||
struct JobData $C.mapname("xce_jobs_data_t") {
|
||
ntotal @0: UInt32 $C.mapname("n_total");
|
||
pos @1: UInt32;
|
||
jobs @2: List(Job) $C.maplistcount("count");
|
||
}
|
||
|
||
struct SchedJobData $C.mapname("xce_jobs_data_t") {
|
||
ntotal @0: UInt32 $C.mapname("n_total");
|
||
pos @1: UInt32;
|
||
jobs @2: List(SchedJob) $C.maplistcount("count");
|
||
}
|
||
|
||
struct WorkerData $C.mapname("xce_workers_data_t"){
|
||
ntotal @0: UInt32 $C.mapname("n_total");
|
||
pos @1: UInt32;
|
||
workers @2: List(Worker) $C.maplistcount("count");
|
||
}
|
||
|
||
struct ResourceData $C.mapname("xce_resource_data_t") {
|
||
ntotal @0: UInt32 $C.mapname("n_total");
|
||
pos @1: UInt32;
|
||
resources @2: List(Resource) $C.maplistcount("count");
|
||
}
|
||
|
||
struct ResourceVal $C.mapname("xce_resval_t") {
|
||
name @0: Text $C.mapname("name");
|
||
host @1: Text $C.mapname("host");
|
||
val @2: Value $C.mapname("val");
|
||
}
|
||
|
||
struct ResourceVals $C.mapname("xce_resval_data_t") {
|
||
ntotal @0: UInt32 $C.mapname("n_total");
|
||
pos @1: UInt32;
|
||
vals @2: List(ResourceVal) $C.mapname("vals") $C.maplistcount("count");
|
||
}
|
||
|
||
struct Data $C.mapname("xce_data_t"){
|
||
data :union $C.mapuniontag("table") {
|
||
schedjobs @0: SchedJobData $C.mapname("jobs");
|
||
schedworkers @1: WorkerData $C.mapname("workers");
|
||
jobs @2: JobData $C.mapname("jobs");
|
||
workers @3: WorkerData $C.mapname("workers");
|
||
resources @4: ResourceData $C.mapname("resources");
|
||
resvals @5: ResourceVals $C.mapname("resvals");
|
||
}
|
||
}
|
||
|
||
struct Get $C.mapname("xce_get_t") {
|
||
table @0: UInt32;
|
||
condition @1: MatchCondition;
|
||
}
|
||
|
||
struct ResValList $C.mapname("xce_resval_list_t") {
|
||
vals @0: List(ResourceVal) $C.mapname("vals") $C.maplistcount("count");
|
||
}
|
||
|
||
struct ResUsage $C.mapname("xce_rusage_t") {
|
||
name @0: Text $C.mapname("resname");
|
||
value @1: Float64 $C.mapname("value");
|
||
}
|
||
|
||
struct JobRusages $C.mapname("xce_job_rusages_t") {
|
||
jobid @0: UInt64 $C.mapname("jobid");
|
||
pids @1: List(Int32) $C.mapname("pids") $C.maplistcount("npids");
|
||
pgids @2: List(Int32) $C.mapname("pgids") $C.maplistcount("npgids");
|
||
usages @3: List(ResUsage) $C.mapname("usages") $C.maplistcount("count");
|
||
}
|
||
|
||
struct Update $C.mapname("xce_update_data_t") {
|
||
data :union $C.mapuniontag("table") {
|
||
nodata @0: Void;
|
||
resvals @1: ResValList $C.mapname("resvals");
|
||
jobrusages @2: JobRusages $C.mapname("jobrusages");
|
||
}
|
||
}
|
||
|
||
struct ResAlloc $C.mapname("xce_alloc_t"){
|
||
name @0: Text $C.mapname("resname");
|
||
resvId @1: Text $C.mapname("reserve");
|
||
instid @2: UInt64;
|
||
value @3: Float64;
|
||
from @4: UInt64; # alloc start time
|
||
duration @5: Int64;
|
||
}
|
||
|
||
struct HostAlloc $C.mapname("xce_host_alloc_t"){
|
||
host @0: Text;
|
||
allocs @1: List(ResAlloc) $C.mapname("allocs") $C.maplistcount("nallocs");
|
||
}
|
||
|
||
struct JobAlloc $C.mapname("xce_job_alloc_t"){
|
||
jobid @0: UInt64;
|
||
allocs @1: List(HostAlloc) $C.mapname("allocs") $C.maplistcount("nhosts");
|
||
allocsShared @2: List(ResAlloc) $C.mapname("allocs_shared") $C.maplistcount("nres_shared");
|
||
}
|
||
|
||
struct JobDispatch $C.mapname("xce_job_dispatch_t"){
|
||
allocs @0: List(JobAlloc) $C.maplistcount("njobs");
|
||
}
|
||
|
||
struct HostLaunch $C.mapname("xce_host_launch_t") {
|
||
host @0: Text $C.mapname("host");
|
||
nranks @1: UInt32 $C.mapname("nranks");
|
||
}
|
||
|
||
struct JobLaunch $C.mapname("xce_launch_t") {
|
||
hosts @0: List(HostLaunch) $C.mapname("hosts") $C.maplistcount("nhosts");
|
||
}
|
||
|
||
struct JobExecInfo $C.mapname("xce_job_exec_t"){
|
||
jobid @0: UInt64;
|
||
startinfo @1: JobStartInfo $C.mapname("start_info");
|
||
launchinfo @2: JobLaunch $C.mapname("launchinfo");
|
||
}
|
||
|
||
struct JobFinishInfo $C.mapname("xce_jobfinish_t") {
|
||
jobid @0: UInt64;
|
||
code @1: Int32 $C.mapname("exit_code");
|
||
}
|
||
|
||
struct TaskPidInfo $C.mapname("xce_task_pidinfo_t") {
|
||
jobid @0: UInt64 $C.mapname("jobid");
|
||
host @1: Text $C.mapname("host");
|
||
taskidx @2: Int32 $C.mapname("taskidx");
|
||
pids @3: List(UInt32) $C.mapname("pids") $C.maplistcount("npids");
|
||
pgids @4: List(UInt32) $C.mapname("pgids") $C.maplistcount("npgids");
|
||
exdata @5: Text $C.mapname("exdata");
|
||
clientidx @6: UInt32 $C.mapname("clientidx");
|
||
}
|
||
|
||
struct TaskFailInfo $C.mapname("xce_task_failinfo_t") {
|
||
jobid @0: UInt64 $C.mapname("jobid");
|
||
host @1: Text $C.mapname("host");
|
||
taskidx @2: Int32 $C.mapname("taskidx");
|
||
code @3: Int32 $C.mapname("code");
|
||
detail @4: Text $C.mapname("detail");
|
||
clientidx @5: Int32 $C.mapname("clientidx");
|
||
}
|
||
|
||
struct TaskFinish $C.mapname("xce_task_finish_t") {
|
||
jobid @0: UInt64 $C.mapname("jobid");
|
||
host @1: Text $C.mapname("host");
|
||
taskidx @2: UInt32 $C.mapname("taskidx");
|
||
exitcode @3: Int32 $C.mapname("exit_code");
|
||
clientidx @4: Int32 $C.mapname("clientidx");
|
||
}
|
||
|
||
struct NotifyData $C.mapname("xce_notify_data_t"){
|
||
data :union $C.mapuniontag("event") $C.mapname("data"){
|
||
jobPidInfo @0: JobPIDInfo $C.mapname("pidinfo");
|
||
jobFailInfo @1: JobFail $C.mapname("failinfo");
|
||
jobDispatchFail @2: JobFail $C.mapname("dispatchfail");
|
||
jobFinishInfo @3: JobFinishInfo $C.mapname("finishinfo");
|
||
schedStartJob @4: UInt64 $C.mapname("sched_start_job");
|
||
schedExitJob @5: UInt64 $C.mapname("sched_exit_job");
|
||
schedFailJob @6: UInt64 $C.mapname("sched_fail_job");
|
||
schedUpdateJob @7: SchedJob $C.mapname("sched_update_job");
|
||
schedUpdateResvals @8: ResValList $C.mapname("sched_update_resvals");
|
||
stoppedJob @9: UInt64 $C.mapname("stopped_job");
|
||
resumedJob @10: UInt64 $C.mapname("resumed_job");
|
||
ckptinfo @11: CheckpointInfo $C.mapname("ckptinfo");
|
||
schedUpdateQueue @12: QueueConf $C.mapname("sched_update_queue");
|
||
taskPidInfo @13: TaskPidInfo $C.mapname("taskinfo");
|
||
taskFailInfo @14: TaskFailInfo $C.mapname("taskfailinfo");
|
||
taskFinish @15: TaskFinish $C.mapname("taskfinish");
|
||
postDoneJob @16: UInt64 $C.mapname("pdone_job");
|
||
schedJobRusages @17: SchedJobRusages $C.mapname("sched_job_rusages");
|
||
updateJob @18: Job $C.mapname("update_job");
|
||
}
|
||
}
|
||
|
||
struct CtrlPeer $C.mapname("xce_ctrlpeer_t") {
|
||
name @0: Text;
|
||
host @1: Text;
|
||
port @2: UInt32;
|
||
}
|
||
|
||
struct CtrlStatus $C.mapname("xce_ctrlstatus_t") {
|
||
clusterName @0: Text;
|
||
primaryName @1: Text;
|
||
responderName @2: Text;
|
||
responderRole @3: UInt8;
|
||
term @4: UInt64;
|
||
configVersion @5: UInt64;
|
||
configChecksum @6: Text;
|
||
committedConfigVersion @7: UInt64;
|
||
committedLsn @8: UInt64;
|
||
appliedLsn @9: UInt64;
|
||
fenced @10: Bool;
|
||
peers @11: List(CtrlPeer) $C.mapname("peers.list") $C.maplistcount("peers.count");
|
||
peerCount @12: UInt32;
|
||
responderTimeMs @13: UInt64 $C.mapname("responder_time_ms"); # responder's wall-clock at sample time (epoch milliseconds)
|
||
}
|
||
|
||
struct ClusterConf $C.mapname("xce_clusterconf_t") {
|
||
name @0: Text;
|
||
resources @1: List(Resource) $C.mapname("reslist.res") $C.maplistcount("reslist.count");
|
||
queues @2: List(QueueConf) $C.mapname("qlist.queues") $C.maplistcount("qlist.count");
|
||
workers @3: List(Worker) $C.mapname("wlist.workers") $C.maplistcount("wlist.count");
|
||
}
|
||
|
||
struct JobSubmitData $C.mapname("job_submit_data_t") {
|
||
queue @0: Text;
|
||
user @1: Text;
|
||
mintasks @2: UInt32;
|
||
maxtasks @3: UInt32;
|
||
jobname @4: Text $C.mapname("job_name");
|
||
project @5: Text $C.mapname("project_name");
|
||
ugroup @6: Text $C.mapname("user_group");
|
||
description @7: Text $C.mapname("job_description");
|
||
resreq @8: Text;
|
||
startinfo @9: JobStartInfo $C.mapname("start_info");
|
||
}
|
||
|
||
struct JobSubmit $C.mapname("xce_job_submit_t") {
|
||
jobid @0: UInt64;
|
||
arraySize @1: UInt64 $C.mapname("array_size");
|
||
submit @2: JobSubmitData;
|
||
}
|
||
|
||
struct JobSignal $C.mapname("xce_signal_t") {
|
||
signo @0: UInt32 $C.mapname("signo");
|
||
force @1: Int32 $C.mapname("force");
|
||
}
|
||
|
||
struct CheckpointOptions $C.mapname("xce_ckpt_options_t"){
|
||
kill @0: UInt32 $C.mapname("kill");
|
||
method @1: Text $C.mapname("method");
|
||
}
|
||
|
||
struct CheckpointInfo $C.mapname("xce_ckptinfo_t") {
|
||
jobid @0: UInt64;
|
||
ckptinfo @1: Text;
|
||
}
|
||
|
||
struct JobControl $C.mapname("xce_jobctrl_t") {
|
||
jobid @0: UInt64 $C.mapname("jobid");
|
||
u :union $C.mapuniontag("action") $C.mapname("u") {
|
||
siginfo @1: JobSignal;
|
||
ckptoption @2: CheckpointOptions;
|
||
}
|
||
}
|
||
|
||
struct QueueControl $C.mapname("xce_queuectrl_t") {
|
||
action @0: UInt32 $C.mapname("action");
|
||
queue @1: Text $C.mapname("queue");
|
||
}
|
||
|
||
struct HostControl $C.mapname("xce_hostctrl_t") {
|
||
action @0: UInt32 $C.mapname("action");
|
||
hostname @1: Text $C.mapname("hostname");
|
||
}
|
||
|
||
struct RequestReply $C.mapname("xce_reply_t"){
|
||
seq @0: UInt64;
|
||
opcode @1: OPCode;
|
||
status @2: Int32;
|
||
|
||
data : union {
|
||
nodata @3: Void;
|
||
debugmsg @4: Text;
|
||
getdata @5: Data;
|
||
jobid @6: UInt64;
|
||
clusterconf @7: ClusterConf;
|
||
replconfigreplymsg @8: ReplConfigReply;
|
||
replconfigpushreplymsg @9: ReplConfigPushReply;
|
||
upgradeconfigreplymsg @10: UpgradeConfigReply;
|
||
ctrlhelloreplymsg @13: CtrlHelloReply;
|
||
requestvotereplymsg @14: RequestVoteReply;
|
||
stateReportMsg @15: StateReport $C.mapname("state_report"); # heartbeat reply: standby state
|
||
walackreplymsg @16: WalAckReply $C.mapname("walackreplymsg"); # WAL apply ACK from standby
|
||
ctrlstatus @17: CtrlStatus; # ctrlstatus reply: identity/health snapshot
|
||
}
|
||
|
||
redirectHost @11: Text $C.mapname("redirect_host");
|
||
redirectPort @12: UInt16 $C.mapname("redirect_port");
|
||
}
|
||
|
||
struct TaskHost $C.mapname("xce_taskhost_t") {
|
||
host @0: Text;
|
||
ntasks @1: UInt32;
|
||
exechome @2: Text;
|
||
execcwd @3: Text;
|
||
execuser @4: Text;
|
||
cmdline @5: Text;
|
||
exdata @6: Text;
|
||
launchtype @7: Text $C.mapname("launch_type");
|
||
}
|
||
|
||
struct RunTask $C.mapname("xce_taskrun_t") {
|
||
clientidx @0: Int32 $C.mapname("clientidx");
|
||
hosts @1: List(TaskHost) $C.mapname("taskhosts") $C.maplistcount("nhosts");
|
||
jobid @2: UInt64 $C.mapname("jobid");
|
||
}
|
||
|
||
struct PendReason $C.mapname("xce_pendreason_t") {
|
||
jobid @0: UInt64 $C.mapname("jobid");
|
||
reasons @1: JobPendReason $C.mapname("reasons");
|
||
}
|
||
|
||
struct PendReasonSet $C.mapname("xce_preason_set_t") {
|
||
reason @0: List(PendReason) $C.mapname("reasons") $C.maplistcount("nreasons");
|
||
}
|
||
|
||
struct ClearPReason $C.mapname("xce_clear_reason_t") {
|
||
jobs @0: List(UInt64) $C.mapname("jobs") $C.maplistcount("njobs");
|
||
}
|
||
|
||
struct Request $C.mapname("xce_request_t"){
|
||
opcode @0: OPCode;
|
||
seq @1: UInt64;
|
||
from @2: PeerInfo;
|
||
body : union {
|
||
nomsg @3: Void;
|
||
replymsg @4: RequestReply;
|
||
hellomsg @5: Text;
|
||
joinmsg @6: JoinData;
|
||
configpushmsg @7: ConfigData;
|
||
pushmsg @8: Data;
|
||
getmsg @9: Get;
|
||
datamsg @10: Dataset;
|
||
dispatchmsg @11: JobDispatch;
|
||
jobstartmsg @12: JobExecInfo;
|
||
notifymsg @13: NotifyData;
|
||
jobsubmsg @14: JobSubmit;
|
||
updatemsg @15: Update;
|
||
jobctrlmsg @16: JobControl;
|
||
queuectrlmsg @17: QueueControl;
|
||
runtaskmsg @18: RunTask;
|
||
preasonmsg @19: PendReasonSet;
|
||
clrpreasonmsg @20: ClearPReason;
|
||
hostctrlmsg @21: HostControl;
|
||
replconfigrequestmsg @22: ReplConfigRequest;
|
||
replconfigpushrequestmsg @23: ReplConfigPushRequest;
|
||
upgradeconfigrequestmsg @24: UpgradeConfigRequest;
|
||
heartbeatmsg @25: HeartbeatMsg;
|
||
ctrlhellomsg @26: CtrlHelloRequest;
|
||
requestvotemsg @27: RequestVoteRequest;
|
||
walcatchupentrymsg @28: WalCatchupEntry;
|
||
snapshottransfermsg @29: SnapshotTransfer;
|
||
}
|
||
}
|
||
|
||
struct HashZArrayConfig $C.mapname("hash_zarray_config_t") {
|
||
nHashSlots @0: UInt32 $C.mapname("nhash_slots");
|
||
delta @1: UInt32 $C.mapname("delta");
|
||
}
|
||
|
||
struct ResmgrConfig $C.mapname("resmgr_config_t") {
|
||
nHashSlots @0: UInt32 $C.mapname("nhash_slots");
|
||
delta @1: UInt32 $C.mapname("delta");
|
||
instDelta @2: UInt32 $C.mapname("inst_delta");
|
||
}
|
||
|
||
struct MasterCandidate $C.mapname("master_candidate_t") {
|
||
name @0: Text $C.mapname("name");
|
||
host @1: Text $C.mapname("host");
|
||
port @2: UInt32 $C.mapname("port");
|
||
}
|
||
|
||
struct HostConfig $C.mapname("xce_hostconfig_t") {
|
||
name @0: Text $C.mapname("name");
|
||
model @1: Text $C.mapname("hostmodel");
|
||
type @2: Text $C.mapname("hosttype");
|
||
mxj @3: UInt32 $C.mapname("mxj");
|
||
resources @4: List(ConfigVar) $C.mapname("resources") $C.maplistcount("nresources");
|
||
thresholds @5: List(Threshold) $C.mapname("thresholds") $C.maplistcount("nthresholds");
|
||
dispatchwindows @6: TimeWindow $C.mapname("dispatchwindow");
|
||
}
|
||
|
||
struct ScalerConfig $C.mapname("scaler_config_t") {
|
||
name @0: Text $C.mapname("name");
|
||
type @1: Text $C.mapname("prov_type");
|
||
confPath @2: Text $C.mapname("conf_path");
|
||
scriptPath @3: Text $C.mapname("script_path");
|
||
scriptOption @4: Text $C.mapname("script_option");
|
||
}
|
||
|
||
struct LimitConfig $C.mapname("xce_limit_config_t") {
|
||
name @0: Text $C.mapname("name");
|
||
users @1: GroupConfig $C.mapname("users");
|
||
queues @2: GroupConfig $C.mapname("queues");
|
||
projects @3: GroupConfig $C.mapname("projects");
|
||
hosts @4: GroupConfig $C.mapname("hosts");
|
||
val @5: LimitValue $C.mapname("val");
|
||
perUser @6: Bool $C.mapname("per_user");
|
||
perQueue @7: Bool $C.mapname("per_queue");
|
||
perProject @8: Bool $C.mapname("per_project");
|
||
perHost @9: Bool $C.mapname("per_host");
|
||
}
|
||
|
||
struct ClusterConfig $C.mapname("cluster_config_t") {
|
||
clusterName @0: Text $C.mapname("clustername");
|
||
port @1: UInt32 $C.mapname("port");
|
||
candidates @2: List(MasterCandidate) $C.mapname("candidates") $C.maplistcount("ncandidates");
|
||
heartbeatInterval @3: UInt32 $C.mapname("heartbeat_interval");
|
||
inactiveHeartbeatCount @4: UInt32 $C.mapname("inactive_heartbeat_count");
|
||
fetchJobInterval @5: UInt32 $C.mapname("fetch_job_interval");
|
||
netioTimeout @6: UInt32 $C.mapname("netio_timeout");
|
||
vcpus @7: UInt32 $C.mapname("vcpus"); # deprecated: use --vcpus CLI flag; kept for config blob backward compatibility
|
||
resources @8: List(Resource) $C.mapname("resources") $C.maplistcount("nres");
|
||
queues @9: List(QueueConf) $C.mapname("queues") $C.maplistcount("nqueues");
|
||
ugroups @10: List(GroupConfig) $C.mapname("ugroups") $C.maplistcount("nugroups");
|
||
hgroups @11: List(GroupConfig) $C.mapname("hgroups") $C.maplistcount("nhgroups");
|
||
hostConfigs @12: List(HostConfig) $C.mapname("hconfigs") $C.maplistcount("nhconfigs");
|
||
jobmgr @13: HashZArrayConfig;
|
||
workermgr @14: HashZArrayConfig;
|
||
qmgr @15: HashZArrayConfig;
|
||
resmgr @16: ResmgrConfig;
|
||
xstrings @17: HashZArrayConfig;
|
||
hproviders @18: List(ScalerConfig) $C.mapname("hproviders") $C.maplistcount("nhproviders");
|
||
limits @19: List(LimitConfig) $C.mapname("limits") $C.maplistcount("nlimits");
|
||
configGeneration @20: UInt64 $C.mapname("config_generation");
|
||
}
|
||
|
||
# ── Persistence configuration (Part 2.6) ────────────────────────────
|
||
# $C.mapname uses capnp_persist_cfg_t (not persist_config_t) to avoid
|
||
# collision with the plain-C struct already defined in persist.h.
|
||
struct PersistConfig $C.mapname("capnp_persist_cfg_t") {
|
||
walSyncMs @0 :UInt32 = 10;
|
||
snapIntervalS @1 :UInt32 = 60;
|
||
snapWalSizeMb @2 :UInt32 = 64;
|
||
finishedRetainS @3 :UInt32 = 3600;
|
||
workerRetainS @4 :UInt32 = 86400;
|
||
syncTimeoutMs @5 :UInt32 = 50;
|
||
snapKeepN @6 :UInt32 = 2;
|
||
transientJobTimeoutS @7 :UInt32 = 300;
|
||
}
|
||
|
||
# ── Snapshot header (Part 2.5) ───────────────────────────────────────
|
||
struct SnapshotHeader $C.mapname("capnp_snap_hdr_t") {
|
||
lsn @0 :UInt64;
|
||
term @1 :UInt64;
|
||
timestamp @2 :UInt64;
|
||
version @3 :UInt32 = 1;
|
||
nJobs @4 :UInt32;
|
||
nQueues @5 :UInt32;
|
||
nWorkers @6 :UInt32;
|
||
checksum @7 :UInt32;
|
||
headerChecksum @8 :UInt32;
|
||
}
|
||
|
||
# ── HA heartbeat ────────────────────────────────────────────────────
|
||
struct HeartbeatMsg $C.mapname("heartbeat_msg_t") {
|
||
senderPriority @0: UInt32 $C.mapname("sender_priority");
|
||
term @1: UInt64 $C.mapname("term");
|
||
committedLsn @2: UInt64 $C.mapname("committed_lsn");
|
||
committedConfigTag @3: UInt64 $C.mapname("committed_config_tag");
|
||
}
|
||
|
||
# ── Config replication message structs ──────────────────────────────
|
||
struct ReplConfigRequest $C.mapname("repl_config_req_t") {
|
||
versionTag @0: UInt64 $C.mapname("version_tag");
|
||
}
|
||
|
||
struct ReplConfigReply $C.mapname("repl_config_reply_t") {
|
||
upToDate @0: Bool $C.mapname("up_to_date");
|
||
versionTag @1: UInt64 $C.mapname("version_tag");
|
||
sha256 @2: Data $C.mapname("sha256");
|
||
bytes @3: Data $C.mapname("bytes");
|
||
}
|
||
|
||
struct ReplConfigPushRequest $C.mapname("repl_config_push_req_t") {
|
||
versionTag @0: UInt64 $C.mapname("version_tag");
|
||
sha256 @1: Data $C.mapname("sha256");
|
||
bytes @2: Data $C.mapname("bytes");
|
||
}
|
||
|
||
struct ReplConfigPushReply $C.mapname("repl_config_push_reply_t") {
|
||
accepted @0: Bool $C.mapname("accepted");
|
||
versionTag @1: UInt64 $C.mapname("version_tag");
|
||
}
|
||
|
||
struct UpgradeConfigRequest $C.mapname("upgrade_config_req_t") {
|
||
sha256 @0: Data $C.mapname("sha256");
|
||
bytes @1: Data $C.mapname("bytes");
|
||
}
|
||
|
||
struct UpgradeConfigReply $C.mapname("upgrade_config_reply_t") {
|
||
versionTag @0: UInt64 $C.mapname("version_tag");
|
||
error @1: Text $C.mapname("error");
|
||
ackedCount @2: UInt32 $C.mapname("acked_count");
|
||
}
|
||
|
||
# ── WAL replication envelope (Part 6 — NOT the on-disk format) ───────
|
||
# Named ReplEnvelope to avoid collision with the on-disk wal_entry_hdr_t.
|
||
struct ReplEnvelope $C.mapname("capnp_repl_envelope_t") {
|
||
term @0 :UInt64;
|
||
lsn @1 :UInt64;
|
||
entryType @2 :UInt32;
|
||
payload @3 :Data;
|
||
crc @4 :UInt32;
|
||
prevCrc @5 :UInt32;
|
||
}
|
||
|
||
# ── HA election and replication message structs (Task 1.1) ───────────
|
||
|
||
# Heartbeat reply state report — standby piggybacks its replication state
|
||
# onto every OPCode_nop reply so the primary can drive convergence decisions.
|
||
struct StateReport $C.mapname("state_report_t") {
|
||
appliedLsn @0 :UInt64 $C.mapname("applied_lsn"); # standby's last applied WAL LSN
|
||
configTag @1 :UInt64 $C.mapname("config_tag"); # standby's current config version tag
|
||
}
|
||
|
||
struct RequestVoteRequest $C.mapname("request_vote_req_t") {
|
||
term @0 :UInt64 $C.mapname("term");
|
||
candidateId @1 :Text $C.mapname("candidate_id");
|
||
configTag @2 :UInt64 $C.mapname("config_tag");
|
||
lastLogLsn @3 :UInt64 $C.mapname("last_log_lsn");
|
||
}
|
||
|
||
struct RequestVoteReply $C.mapname("request_vote_reply_t") {
|
||
term @0 :UInt64 $C.mapname("term");
|
||
voteGranted @1 :Bool $C.mapname("vote_granted");
|
||
peerConfigTag @2 :UInt64 $C.mapname("peer_config_tag");
|
||
peerLastLogLsn @3 :UInt64 $C.mapname("peer_last_log_lsn");
|
||
}
|
||
|
||
struct ProbeConfigTagRequest $C.mapname("probe_config_tag_req_t") {
|
||
name @0 :Text $C.mapname("name");
|
||
}
|
||
|
||
struct ProbeConfigTagReply $C.mapname("probe_config_tag_reply_t") {
|
||
configTag @0 :UInt64 $C.mapname("config_tag");
|
||
}
|
||
|
||
struct WalAckRequest $C.mapname("wal_ack_req_t") {
|
||
term @0 :UInt64 $C.mapname("term");
|
||
lsn @1 :UInt64 $C.mapname("lsn");
|
||
}
|
||
|
||
struct WalAckReply $C.mapname("wal_ack_reply_t") {
|
||
term @0 :UInt64 $C.mapname("term");
|
||
lsn @1 :UInt64 $C.mapname("lsn");
|
||
success @2 :Bool $C.mapname("success");
|
||
}
|
||
|
||
struct UpgradeConfigAck $C.mapname("upgrade_config_ack_t") {
|
||
term @0 :UInt64 $C.mapname("term");
|
||
ackedTag @1 :UInt64 $C.mapname("acked_tag");
|
||
success @2 :Bool $C.mapname("success");
|
||
}
|
||
|
||
# ── Snapshot chunk transfer (Req 4.1–4.2) ───────────────────────────
|
||
# Used for OPCode_snapshotTransfer: full state snapshot transferred as
|
||
# a sequence of fixed-size chunks (64 KB max each). Restart-on-reconnect
|
||
# semantics — no resumption state needed. Fields @0 and @1 are preserved
|
||
# for wire-protocol continuity; @2–@5 carry the chunk envelope.
|
||
struct SnapshotTransfer $C.mapname("snapshot_transfer_t") {
|
||
snapshotLsn @0 :UInt64 $C.mapname("snapshot_lsn"); # checkpoint LSN this snapshot corresponds to
|
||
payload @1 :Data $C.mapname("payload"); # chunk payload (64 KB max)
|
||
chunkIndex @2 :UInt32 $C.mapname("chunk_index"); # 0-based chunk number
|
||
chunkOffset @3 :UInt64 $C.mapname("chunk_offset"); # byte offset in the full snapshot
|
||
totalSize @4 :UInt64 $C.mapname("total_size"); # total snapshot size in bytes
|
||
isLast @5 :Bool $C.mapname("is_last"); # true for the final chunk
|
||
}
|
||
|
||
struct WalCatchupEntry $C.mapname("wal_catchup_entry_t") {
|
||
lsn @0 :UInt64 $C.mapname("lsn");
|
||
entryType @1 :UInt8 $C.mapname("entry_type");
|
||
payload @2 :Data $C.mapname("payload");
|
||
entries @3 :List(WalCatchupEntry) $C.mapname("entries") $C.maplistcount("n_entries");
|
||
}
|
||
|
||
struct CtrlHelloRequest $C.mapname("ctrl_hello_req_t") {
|
||
name @0 :Text $C.mapname("name");
|
||
key @1 :Data $C.mapname("key");
|
||
appliedLsn @2 :UInt64 $C.mapname("applied_lsn");
|
||
configTag @3 :UInt64 $C.mapname("config_tag");
|
||
}
|
||
|
||
struct CtrlHelloReply $C.mapname("ctrl_hello_reply_t") {
|
||
committedLsn @0 :UInt64 $C.mapname("committed_lsn");
|
||
needsSnapshot @1 :Bool $C.mapname("needs_snapshot");
|
||
configTag @2 :UInt64 $C.mapname("config_tag");
|
||
}
|