This is an automated email from the ASF dual-hosted git repository.
dataroaring pushed a commit to branch branch-2.0
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/branch-2.0 by this push:
new 5f5cb635df3 [chore](tablet scheduler) update sched ctx err #27514
(#31189)
5f5cb635df3 is described below
commit 5f5cb635df332ffc74c22eac38834e3c9c4e2ca7
Author: yujun <yu.jun.re...@gmail.com>
AuthorDate: Tue Apr 30 14:18:04 2024 +0800
[chore](tablet scheduler) update sched ctx err #27514 (#31189)
---
.../org/apache/doris/clone/BeLoadRebalancer.java | 6 ++---
.../org/apache/doris/clone/SchedException.java | 1 +
.../org/apache/doris/clone/TabletSchedCtx.java | 12 +++++++++
.../org/apache/doris/clone/TabletScheduler.java | 30 ++++++++++++----------
4 files changed, 33 insertions(+), 16 deletions(-)
diff --git
a/fe/fe-core/src/main/java/org/apache/doris/clone/BeLoadRebalancer.java
b/fe/fe-core/src/main/java/org/apache/doris/clone/BeLoadRebalancer.java
index 93d884af5ce..c81271eb890 100644
--- a/fe/fe-core/src/main/java/org/apache/doris/clone/BeLoadRebalancer.java
+++ b/fe/fe-core/src/main/java/org/apache/doris/clone/BeLoadRebalancer.java
@@ -288,7 +288,7 @@ public class BeLoadRebalancer extends Rebalancer {
}
}
if (!setSource) {
- throw new SchedException(Status.UNRECOVERABLE, "unable to take src
slot");
+ throw new SchedException(Status.UNRECOVERABLE,
SubCode.DIAGNOSE_IGNORE, "unable to take src backend slot");
}
// Select a low load backend as destination.
@@ -331,7 +331,7 @@ public class BeLoadRebalancer extends Rebalancer {
}
if (candidates.isEmpty()) {
- throw new SchedException(Status.UNRECOVERABLE, "unable to find low
backend");
+ throw new SchedException(Status.UNRECOVERABLE,
SubCode.DIAGNOSE_IGNORE, "unable to find low dest backend");
}
List<BePathLoadStatPair> candFitPaths = Lists.newArrayList();
@@ -368,7 +368,7 @@ public class BeLoadRebalancer extends Rebalancer {
}
throw new SchedException(Status.SCHEDULE_FAILED, SubCode.WAITING_SLOT,
- "unable to find low backend");
+ "beload waiting for dest backend slot");
}
}
diff --git
a/fe/fe-core/src/main/java/org/apache/doris/clone/SchedException.java
b/fe/fe-core/src/main/java/org/apache/doris/clone/SchedException.java
index a343e6543c3..cb7080299ec 100644
--- a/fe/fe-core/src/main/java/org/apache/doris/clone/SchedException.java
+++ b/fe/fe-core/src/main/java/org/apache/doris/clone/SchedException.java
@@ -31,6 +31,7 @@ public class SchedException extends Exception {
NONE,
WAITING_DECOMMISSION,
WAITING_SLOT,
+ DIAGNOSE_IGNORE,
}
private Status status;
diff --git
a/fe/fe-core/src/main/java/org/apache/doris/clone/TabletSchedCtx.java
b/fe/fe-core/src/main/java/org/apache/doris/clone/TabletSchedCtx.java
index b9606f09d0a..912fc1bb316 100644
--- a/fe/fe-core/src/main/java/org/apache/doris/clone/TabletSchedCtx.java
+++ b/fe/fe-core/src/main/java/org/apache/doris/clone/TabletSchedCtx.java
@@ -424,6 +424,18 @@ public class TabletSchedCtx implements
Comparable<TabletSchedCtx> {
this.errMsg = errMsg;
}
+ public String getErrMsg() {
+ return errMsg;
+ }
+
+ public SubCode getSchedFailedCode() {
+ return schedFailedCode;
+ }
+
+ public void setSchedFailedCode(SubCode code) {
+ schedFailedCode = code;
+ }
+
public CloneTask getCloneTask() {
return cloneTask;
}
diff --git
a/fe/fe-core/src/main/java/org/apache/doris/clone/TabletScheduler.java
b/fe/fe-core/src/main/java/org/apache/doris/clone/TabletScheduler.java
index ede3e541675..4a22878424d 100644
--- a/fe/fe-core/src/main/java/org/apache/doris/clone/TabletScheduler.java
+++ b/fe/fe-core/src/main/java/org/apache/doris/clone/TabletScheduler.java
@@ -390,15 +390,6 @@ public class TabletScheduler extends MasterDaemon {
AgentBatchTask batchTask = new AgentBatchTask();
for (TabletSchedCtx tabletCtx : currentBatch) {
try {
- if (Config.disable_tablet_scheduler) {
- // do not schedule more tablet is tablet scheduler is
disabled.
- throw new SchedException(Status.FINISHED, "tablet
scheduler is disabled");
- }
- if (Config.disable_balance && tabletCtx.getType() ==
Type.BALANCE) {
- finalizeTabletCtx(tabletCtx,
TabletSchedCtx.State.CANCELLED, Status.UNRECOVERABLE,
- "config disable balance");
- continue;
- }
scheduleTablet(tabletCtx, batchTask);
} catch (SchedException e) {
tabletCtx.setErrMsg(e.getMessage());
@@ -422,6 +413,7 @@ public class TabletScheduler extends MasterDaemon {
Preconditions.checkState(e.getStatus() ==
Status.UNRECOVERABLE, e.getStatus());
// discard
stat.counterTabletScheduledDiscard.incrementAndGet();
+ tabletCtx.setSchedFailedCode(e.getSubCode());
finalizeTabletCtx(tabletCtx,
TabletSchedCtx.State.CANCELLED, e.getStatus(), e.getMessage());
}
continue;
@@ -429,6 +421,8 @@ public class TabletScheduler extends MasterDaemon {
LOG.warn("got unexpected exception, discard this schedule.
tablet: {}",
tabletCtx.getTabletId(), e);
stat.counterTabletScheduledFailed.incrementAndGet();
+ tabletCtx.setSchedFailedCode(SubCode.NONE);
+ tabletCtx.setErrMsg(e.getMessage());
finalizeTabletCtx(tabletCtx, TabletSchedCtx.State.UNEXPECTED,
Status.UNRECOVERABLE, e.getMessage());
continue;
}
@@ -470,6 +464,14 @@ public class TabletScheduler extends MasterDaemon {
* Try to schedule a single tablet.
*/
private void scheduleTablet(TabletSchedCtx tabletCtx, AgentBatchTask
batchTask) throws SchedException {
+ if (Config.disable_tablet_scheduler) {
+ // do not schedule more tablet is tablet scheduler is disabled.
+ throw new SchedException(Status.UNRECOVERABLE,
SubCode.DIAGNOSE_IGNORE, "tablet scheduler is disabled");
+ }
+ if (Config.disable_balance && tabletCtx.getType() == Type.BALANCE) {
+ throw new SchedException(Status.UNRECOVERABLE,
SubCode.DIAGNOSE_IGNORE, "balance is disabled");
+ }
+
long currentTime = System.currentTimeMillis();
tabletCtx.setLastSchedTime(currentTime);
tabletCtx.setLastVisitedTime(currentTime);
@@ -1447,7 +1449,7 @@ public class TabletScheduler extends MasterDaemon {
if (hasBePath) {
throw new SchedException(Status.SCHEDULE_FAILED,
SubCode.WAITING_SLOT,
- "unable to find dest path which can be fit in");
+ "scheduler waiting for dest backend slot");
} else {
throw new SchedException(Status.UNRECOVERABLE,
"unable to find dest path which can be fit in");
@@ -1647,8 +1649,9 @@ public class TabletScheduler extends MasterDaemon {
updateDestPathHash(tabletCtx);
finalizeTabletCtx(tabletCtx, TabletSchedCtx.State.FINISHED,
Status.FINISHED, "finished");
} else {
- finalizeTabletCtx(tabletCtx, TabletSchedCtx.State.CANCELLED,
Status.UNRECOVERABLE,
- request.getTaskStatus().getErrorMsgs().get(0));
+ String errMsg = request.getTaskStatus().getErrorMsgs().get(0);
+ tabletCtx.setErrMsg(errMsg);
+ finalizeTabletCtx(tabletCtx, TabletSchedCtx.State.CANCELLED,
Status.UNRECOVERABLE, errMsg);
}
return true;
@@ -1776,6 +1779,7 @@ public class TabletScheduler extends MasterDaemon {
// Set "resetReplicaState" to true because
// the timeout task should also be considered as UNRECOVERABLE,
// so need to reset replica state.
+ t.setErrMsg("timeout");
finalizeTabletCtx(t, TabletSchedCtx.State.CANCELLED,
Status.UNRECOVERABLE, "timeout");
stat.counterCloneTaskTimeout.incrementAndGet();
});
@@ -1946,7 +1950,7 @@ public class TabletScheduler extends MasterDaemon {
LOG.debug("path hash is not set.", new Exception());
}
throw new SchedException(Status.SCHEDULE_FAILED,
SubCode.WAITING_SLOT,
- "path hash is not set");
+ "backend " + beId + " path hash is not set");
}
Slot slot = pathSlots.get(pathHash);
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org
For additional commands, e-mail: commits-h...@doris.apache.org
