-
Notifications
You must be signed in to change notification settings - Fork 189
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[JENKINS-55287] Improve error messages for nonresumable Pipelines #363
Changes from 1 commit
ebb5d76
b360ae0
f631a44
686a6e4
eb08b5b
2214c42
b3a6847
2ad7e70
d0c638c
22e7086
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
Original file line number | Diff line number | Diff line change |
---|---|---|
|
@@ -712,17 +712,26 @@ public void onLoad(FlowExecutionOwner owner) throws IOException { | |
this.owner = owner; | ||
|
||
try { | ||
initializeStorage(); // Throws exception and bombs out if we can't load FlowNodes | ||
} catch (Exception ex) { | ||
try { | ||
initializeStorage(); // Throws exception and bombs out if we can't load FlowNodes | ||
} catch (Exception ex) { | ||
LOGGER.log(Level.WARNING, "Error initializing storage and loading nodes, will try to create placeholders for: "+this, ex); | ||
if (!canResume()) { | ||
// This case is expected for PERFORMANCE_OPTIMIZED Pipelines after a Jenkins crash (JENKINS-55287). The specific issue is | ||
// usually that only the FlowStartNode has been persisted and so we cannot find the current head node in storage. | ||
String message = "Unable to resume " + owner.getExecutable() + " because it was not saved before Jenkins shut down. Did Jenkins crash?"; | ||
LOGGER.log(Level.WARNING, message, ex); | ||
ex = new AbortException(message); // Change the exception displayed in the build log to one that is more useful for users. | ||
} else { | ||
LOGGER.log(Level.WARNING, "Unable to resume " + owner.getExecutable() + " because its flow nodes could not be loaded", ex); | ||
} | ||
createPlaceholderNodes(ex); | ||
return; | ||
} catch (Exception ex2) { | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. This is all mostly refactoring. The old code had this structure:
Now it has this structure:
|
||
ex2.addSuppressed(ex); | ||
done = true; | ||
programPromise = Futures.immediateFailedFuture(ex2); | ||
throw new IOException("Failed to even create placeholder nodes for execution", ex2); | ||
} | ||
} catch (Exception ex) { | ||
done = true; | ||
programPromise = Futures.immediateFailedFuture(ex); | ||
throw new IOException("Failed to even create placeholder nodes for execution", ex); | ||
return; | ||
} | ||
|
||
try { | ||
|
@@ -737,12 +746,17 @@ public void onLoad(FlowExecutionOwner owner) throws IOException { | |
} | ||
} else { // See if we can/should resume build | ||
if (canResume()) { | ||
// TODO: I think we need to null out persistedClean here and resave the execution so that future | ||
// resumptions are not misled by the fact that persistedClean was true before this resumption. | ||
// CpsThreadGroup.run nulls out persistedClean after every step but it does not save that value | ||
// for PERFORMANCE_OPTIMIZED Pipelines. | ||
dwnusbaum marked this conversation as resolved.
Show resolved
Hide resolved
|
||
loadProgramAsync(getProgramDataFile()); | ||
} else { | ||
// TODO if possible, consider trying to close out unterminated blocks to keep existing graph history | ||
// That way we can visualize the graph in some error cases. | ||
LOGGER.log(Level.WARNING, "Pipeline state not properly persisted, cannot resume "+owner.getUrl()); | ||
throw new IOException("Cannot resume build -- was not cleanly saved when Jenkins shut down."); | ||
String message = "Unable to resume " + owner.getExecutable() + " because it was not saved before Jenkins shut down. Did Jenkins crash?"; | ||
LOGGER.log(Level.WARNING, message); | ||
throw new IOException(message); | ||
} | ||
} | ||
} catch (Exception e) { // Broad catch ensures that failure to load do NOT nuke the master | ||
|
@@ -1526,10 +1540,6 @@ public boolean isPaused() { | |
return false; | ||
} | ||
|
||
private void setPersistedClean(boolean persistedClean) { // Workaround for some issues with anonymous classes. | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Unused |
||
this.persistedClean = persistedClean; | ||
} | ||
|
||
/** | ||
* Pause or unpause the execution. | ||
* | ||
|
@@ -1951,24 +1961,35 @@ void saveExecutionIfDurable() { | |
* Key note: to avoid deadlocks we need to ensure that we don't hold a lock on this CpsFlowExecution when running saveOwner | ||
* or pre-emptively lock the run before locking the execution and saving. */ | ||
void saveOwner() { | ||
/** | ||
* We only want to update persistedClean if the program is over (in which case it doesn't matter), or if Jenkins | ||
* is about to shut down (in which case it controls whether PERFORMANCE_OPTIMIZED Pipelines can be resumed). | ||
*/ | ||
boolean updatePersistedClean = done || Jenkins.get().isTerminating(); | ||
dwnusbaum marked this conversation as resolved.
Show resolved
Hide resolved
|
||
try { | ||
if (this.owner != null && this.owner.getExecutable() instanceof Saveable) { // Null-check covers some anomalous cases we've seen | ||
Saveable saveable = (Saveable)(this.owner.getExecutable()); | ||
persistedClean = true; | ||
if (updatePersistedClean) { | ||
persistedClean = true; | ||
} | ||
if (storage != null && storage.delegate != null) { | ||
// Defensively flush FlowNodes to storage | ||
try { | ||
storage.flush(); | ||
} catch (Exception ex) { | ||
LOGGER.log(Level.WARNING, "Error persisting FlowNodes for execution "+owner, ex); | ||
persistedClean = false; | ||
if (updatePersistedClean) { | ||
persistedClean = false; | ||
} | ||
} | ||
} | ||
saveable.save(); | ||
} | ||
} catch (IOException ex) { | ||
LOGGER.log(Level.WARNING, "Error persisting Run "+owner, ex); | ||
persistedClean = false; | ||
if (updatePersistedClean) { | ||
persistedClean = false; | ||
} | ||
} | ||
} | ||
|
||
|
@@ -2031,7 +2052,12 @@ public void onFailure(Throwable t) { | |
persistOk=false; | ||
LOGGER.log(Level.WARNING, "Error persisting FlowNode storage before shutdown", ioe); | ||
} | ||
persistedClean = persistOk; | ||
if (Jenkins.get().isTerminating()) { | ||
// Only modify persistedClean if Jenkins is shutting down. | ||
// TODO: We could set persistedClean to true when the build is just being paused, but then we would need | ||
// to null it out and save again when the build was unpaused. | ||
dwnusbaum marked this conversation as resolved.
Show resolved
Hide resolved
|
||
persistedClean = persistOk; | ||
} | ||
try { | ||
saveOwner(); | ||
} catch (Exception ex) { | ||
|
Original file line number | Diff line number | Diff line change |
---|---|---|
|
@@ -16,7 +16,6 @@ | |
import org.jenkinsci.plugins.workflow.support.steps.input.InputStepExecution; | ||
import org.junit.Assert; | ||
import org.junit.ClassRule; | ||
import org.junit.Ignore; | ||
import org.junit.Rule; | ||
import org.junit.Test; | ||
import org.jvnet.hudson.test.BuildWatcher; | ||
|
@@ -244,7 +243,6 @@ public void inProgressNormal() throws Exception { | |
} | ||
|
||
@Test | ||
@Ignore | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. These tests have been ignored since they were added. I am not sure why, but I went ahead and unignored them. There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. You probably already knew this, but I just noticed that these test cases seem to be duplicated in There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Yeah, I think I will probably delete the copies in workflow-job after this. Ideally we would run the PCT in the CI builds here with just the core Pipeline plugins so that changes are tested against the newest versions of everything. |
||
public void inProgressMaxPerfCleanShutdown() throws Exception { | ||
final int[] build = new int[1]; | ||
story.then( j -> { | ||
|
@@ -263,14 +261,14 @@ public void inProgressMaxPerfCleanShutdown() throws Exception { | |
}); | ||
} | ||
|
||
@Issue("JENKINS-55287") | ||
@Test | ||
@Ignore | ||
public void inProgressMaxPerfDirtyShutdown() throws Exception { | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. This is the exact situation reported in JENKINS-55287 as far as I can tell. |
||
final int[] build = new int[1]; | ||
final String[] finalNodeId = new String[1]; | ||
story.thenWithHardShutdown( j -> { | ||
runBasicPauseOnInput(j, DEFAULT_JOBNAME, build, FlowDurabilityHint.PERFORMANCE_OPTIMIZED); | ||
// SHOULD still save at end via persist-at-shutdown hooks | ||
// Will not be saved since the persist-at-shutdown hooks will occur after the snapshot for thenWithHardShutdown is taken. | ||
}); | ||
story.then( j->{ | ||
WorkflowJob r = j.jenkins.getItemByFullName(DEFAULT_JOBNAME, WorkflowJob.class); | ||
|
@@ -279,6 +277,7 @@ public void inProgressMaxPerfDirtyShutdown() throws Exception { | |
j.waitForCompletion(run); | ||
assertCompletedCleanly(run); | ||
Assert.assertEquals(Result.FAILURE, run.getResult()); | ||
j.assertLogContains("was not saved before Jenkins shut down", run); | ||
finalNodeId[0] = run.getExecution().getCurrentHeads().get(0).getId(); | ||
}); | ||
story.then(j-> { | ||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Alternatively, we could check this even before we attempt to initialize storage and just bail out right away. I don't know if that is too pessimistic, maybe there is a case where the Pipeline is not resumable but flow nodes are all up to date and so the optimistic behavior here is useful because you will still be able to see the existing flow graph in those cases?