-
Notifications
You must be signed in to change notification settings - Fork 0
Update pauser to handle exception and run unpause as long as possible #33
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
Changes from all commits
92ebb3a
07d8325
08312b2
34a03fc
25c10f6
4d22a34
42f1a30
1d17af9
f087077
d3b54c0
55ba29c
891beb8
c233242
744da27
c413386
600e7b1
9e9132d
909e596
f9dc445
5cca44c
639e132
b93c054
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 |
---|---|---|
@@ -0,0 +1,11 @@ | ||
package com.scalar.admin.kubernetes; | ||
|
||
public class GetTargetAfterPauseFailedException extends PauserException { | ||
public GetTargetAfterPauseFailedException(String message) { | ||
super(message); | ||
} | ||
|
||
public GetTargetAfterPauseFailedException(String message, Throwable cause) { | ||
super(message, cause); | ||
} | ||
} |
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,11 @@ | ||
package com.scalar.admin.kubernetes; | ||
|
||
public class PauseFailedException extends PauserException { | ||
public PauseFailedException(String message) { | ||
super(message); | ||
} | ||
|
||
public PauseFailedException(String message, Throwable cause) { | ||
super(message, cause); | ||
} | ||
} |
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,5 +1,6 @@ | ||
package com.scalar.admin.kubernetes; | ||
|
||
import com.google.common.annotations.VisibleForTesting; | ||
import com.google.common.util.concurrent.Uninterruptibles; | ||
import com.scalar.admin.RequestCoordinator; | ||
import io.kubernetes.client.openapi.Configuration; | ||
|
@@ -13,8 +14,6 @@ | |
import java.util.stream.Collectors; | ||
import javax.annotation.Nullable; | ||
import javax.annotation.concurrent.NotThreadSafe; | ||
import org.slf4j.Logger; | ||
import org.slf4j.LoggerFactory; | ||
|
||
/** | ||
* This class implements a pause operation for Scalar product pods in a Kubernetes cluster. The | ||
|
@@ -34,11 +33,38 @@ | |
@NotThreadSafe | ||
public class Pauser { | ||
|
||
private static final int MAX_UNPAUSE_RETRY_COUNT = 3; | ||
@VisibleForTesting static final int MAX_UNPAUSE_RETRY_COUNT = 3; | ||
|
||
private final Logger logger = LoggerFactory.getLogger(Pauser.class); | ||
private final TargetSelector targetSelector; | ||
|
||
@VisibleForTesting | ||
static final String UNPAUSE_ERROR_MESSAGE = | ||
"Unpause operation failed. Scalar products might still be in a paused state. You" | ||
+ " must restart related pods by using the `kubectl rollout restart deployment" | ||
+ " <DEPLOYMENT_NAME>` command to unpause all pods."; | ||
|
||
@VisibleForTesting | ||
static final String PAUSE_ERROR_MESSAGE = | ||
"Pause operation failed. You cannot use the backup that was taken during this pause" | ||
+ " duration. You need to retry the pause operation from the beginning to" | ||
+ " take a backup."; | ||
|
||
@VisibleForTesting | ||
static final String GET_TARGET_AFTER_PAUSE_ERROR_MESSAGE = | ||
"Failed to find the target pods to examine if the targets pods were updated during" | ||
+ " paused."; | ||
|
||
@VisibleForTesting | ||
static final String STATUS_CHECK_ERROR_MESSAGE = | ||
"Status check failed. You cannot use the backup that was taken during this pause" | ||
+ " duration. You need to retry the pause operation from the beginning to" | ||
+ " take a backup."; | ||
|
||
@VisibleForTesting | ||
static final String STATUS_UNMATCHED_ERROR_MESSAGE = | ||
"The target pods were updated during the pause duration. You cannot use the backup that" | ||
+ " was taken during this pause duration."; | ||
|
||
/** | ||
* @param namespace The namespace where the pods are deployed. | ||
* @param helmReleaseName The Helm release name used to deploy the pods. | ||
|
@@ -77,67 +103,191 @@ public PausedDuration pause(int pauseDuration, @Nullable Long maxPauseWaitTime) | |
"pauseDuration is required to be greater than 0 millisecond."); | ||
} | ||
|
||
TargetSnapshot target; | ||
// Get pods and deployment information before pause. | ||
TargetSnapshot targetBeforePause; | ||
try { | ||
target = targetSelector.select(); | ||
targetBeforePause = getTarget(); | ||
} catch (Exception e) { | ||
throw new PauserException("Failed to find the target pods to pause.", e); | ||
} | ||
|
||
RequestCoordinator coordinator = getRequestCoordinator(target); | ||
|
||
coordinator.pause(true, maxPauseWaitTime); | ||
|
||
Instant startTime = Instant.now(); | ||
// Get RequestCoordinator of Scalar Admin to pause. | ||
RequestCoordinator requestCoordinator; | ||
try { | ||
requestCoordinator = getRequestCoordinator(targetBeforePause); | ||
} catch (Exception e) { | ||
throw new PauserException("Failed to initialize the request coordinator.", e); | ||
} | ||
|
||
Uninterruptibles.sleepUninterruptibly(pauseDuration, TimeUnit.MILLISECONDS); | ||
// From here, we cannot throw exceptions right after they occur because we need to take care of | ||
// the unpause operation failure. We will throw the exception after the unpause operation or at | ||
// the end of this method. | ||
|
||
Instant endTime = Instant.now(); | ||
// Run a pause operation. | ||
PausedDuration pausedDuration = null; | ||
PauseFailedException pauseFailedException = null; | ||
try { | ||
pausedDuration = pauseInternal(requestCoordinator, pauseDuration, maxPauseWaitTime); | ||
} catch (Exception e) { | ||
pauseFailedException = new PauseFailedException(PAUSE_ERROR_MESSAGE, e); | ||
} | ||
|
||
unpauseWithRetry(coordinator, MAX_UNPAUSE_RETRY_COUNT, target); | ||
// Run an unpause operation. | ||
UnpauseFailedException unpauseFailedException = null; | ||
try { | ||
unpauseWithRetry(requestCoordinator, MAX_UNPAUSE_RETRY_COUNT); | ||
} catch (Exception e) { | ||
unpauseFailedException = new UnpauseFailedException(UNPAUSE_ERROR_MESSAGE, e); | ||
} | ||
|
||
TargetSnapshot targetAfterPause; | ||
// Get pods and deployment information after pause. | ||
TargetSnapshot targetAfterPause = null; | ||
GetTargetAfterPauseFailedException getTargetAfterPauseFailedException = null; | ||
try { | ||
targetAfterPause = targetSelector.select(); | ||
targetAfterPause = getTarget(); | ||
} catch (Exception e) { | ||
throw new PauserException( | ||
"Failed to find the target pods to examine if the targets pods were updated during" | ||
+ " paused.", | ||
e); | ||
getTargetAfterPauseFailedException = | ||
new GetTargetAfterPauseFailedException(GET_TARGET_AFTER_PAUSE_ERROR_MESSAGE, e); | ||
} | ||
|
||
if (!target.getStatus().equals(targetAfterPause.getStatus())) { | ||
throw new PauserException("The target pods were updated during paused. Please retry."); | ||
// Check if pods and deployment information are the same between before pause and after pause. | ||
StatusCheckFailedException statusCheckFailedException = null; | ||
StatusUnmatchedException statusUnmatchedException = null; | ||
if (targetAfterPause != null) { | ||
try { | ||
statusUnmatchedException = targetStatusEquals(targetBeforePause, targetAfterPause); | ||
} catch (Exception e) { | ||
statusCheckFailedException = new StatusCheckFailedException(STATUS_CHECK_ERROR_MESSAGE, e); | ||
} | ||
} | ||
|
||
return new PausedDuration(startTime, endTime); | ||
// We use the exceptions as conditions instead of using boolean flags like `isPauseOk`, etc. If | ||
// we use boolean flags, it might cause a bit large number of combinations. For example, if we | ||
// have three flags, they generate 2^3 = 8 combinations. It also might make the nested if | ||
// statement or a lot of branches of the switch statement. That's why we don't use status flags | ||
// for now. | ||
PauserException pauserException = | ||
buildException( | ||
unpauseFailedException, | ||
pauseFailedException, | ||
getTargetAfterPauseFailedException, | ||
statusCheckFailedException, | ||
statusUnmatchedException); | ||
|
||
// Return the final result based on each process. | ||
if (pauserException != null) { | ||
// Some operations failed. | ||
throw pauserException; | ||
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. [minor] Should we have some log messages when Looks like we removed a log in this PR (Sorry if I miss some context here 😓) logger.warn(
"Failed to unpause Scalar product. They are still in paused. You must restart related"
+ " pods by using the `kubectl rollout restart deployment {}`"
+ " command to unpase all pods.",
target.getDeployment().getMetadata().getName());
return; 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. I think we don't need to output such a log message here. I decided to include all messages in the generated exception. Now, we can show users clearer errors by using several kinds of exceptions (e.g., Also, I wanted to see So, there is no problem if we don't have a log here. |
||
} else { | ||
// All operations succeeded. | ||
return pausedDuration; | ||
} | ||
} | ||
|
||
private void unpauseWithRetry( | ||
RequestCoordinator coordinator, int maxRetryCount, TargetSnapshot target) { | ||
@VisibleForTesting | ||
void unpauseWithRetry(RequestCoordinator coordinator, int maxRetryCount) { | ||
int retryCounter = 0; | ||
|
||
while (true) { | ||
try { | ||
coordinator.unpause(); | ||
return; | ||
} catch (Exception e) { | ||
if (++retryCounter >= maxRetryCount) { | ||
logger.warn( | ||
"Failed to unpause Scalar product. They are still in paused. You must restart related" | ||
+ " pods by using the `kubectl rollout restart deployment {}`" | ||
+ " command to unpase all pods.", | ||
target.getDeployment().getMetadata().getName()); | ||
return; | ||
throw e; | ||
} | ||
} | ||
} | ||
} | ||
|
||
@VisibleForTesting | ||
TargetSnapshot getTarget() throws PauserException { | ||
return targetSelector.select(); | ||
} | ||
Comment on lines
+202
to
+205
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. As I mentioned in another comment, I separated this |
||
|
||
@VisibleForTesting | ||
RequestCoordinator getRequestCoordinator(TargetSnapshot target) { | ||
return new RequestCoordinator( | ||
target.getPods().stream() | ||
.map(p -> new InetSocketAddress(p.getStatus().getPodIP(), target.getAdminPort())) | ||
.collect(Collectors.toList())); | ||
} | ||
|
||
@VisibleForTesting | ||
PausedDuration pauseInternal( | ||
RequestCoordinator requestCoordinator, int pauseDuration, @Nullable Long maxPauseWaitTime) { | ||
requestCoordinator.pause(true, maxPauseWaitTime); | ||
Instant startTime = Instant.now(); | ||
Uninterruptibles.sleepUninterruptibly(pauseDuration, TimeUnit.MILLISECONDS); | ||
Instant endTime = Instant.now(); | ||
return new PausedDuration(startTime, endTime); | ||
} | ||
|
||
@VisibleForTesting | ||
@Nullable | ||
StatusUnmatchedException targetStatusEquals(TargetSnapshot before, TargetSnapshot after) { | ||
if (before.getStatus().equals(after.getStatus())) { | ||
return null; | ||
} else { | ||
return new StatusUnmatchedException(STATUS_UNMATCHED_ERROR_MESSAGE); | ||
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. [minor] How about adding 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. Ah, I see. It might have a lot of information that makes the error message very long, but I will check what is included in Thank you for your suggestion! 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. I checked what is included in So, I think it would be better to implement Therefore, I will keep the current implementation as is in this PR, and I will fix it in another PR in the future. |
||
} | ||
} | ||
|
||
@Nullable | ||
@VisibleForTesting | ||
PauserException buildException( | ||
@Nullable UnpauseFailedException unpauseFailedException, | ||
@Nullable PauseFailedException pauseFailedException, | ||
@Nullable GetTargetAfterPauseFailedException getTargetAfterPauseFailedException, | ||
@Nullable StatusCheckFailedException statusCheckFailedException, | ||
@Nullable StatusUnmatchedException statusUnmatchedException) { | ||
PauserException pauserException = null; | ||
|
||
// Treat the unpause failure as most critical because it might cause system unavailability. | ||
if (unpauseFailedException != null) { | ||
pauserException = unpauseFailedException; | ||
} | ||
|
||
// Treat the pause failure as the second priority because the issue might be caused by a | ||
// configuration mistake. | ||
if (pauseFailedException != null) { | ||
if (pauserException == null) { | ||
pauserException = pauseFailedException; | ||
} else { | ||
pauserException.addSuppressed(pauseFailedException); | ||
} | ||
} | ||
|
||
// Treat the getting target failure as the third priority because this issue might be caused by | ||
// a temporary glitch, for example, network failures. | ||
if (getTargetAfterPauseFailedException != null) { | ||
if (pauserException == null) { | ||
pauserException = getTargetAfterPauseFailedException; | ||
} else { | ||
pauserException.addSuppressed(getTargetAfterPauseFailedException); | ||
} | ||
} | ||
|
||
// Treat the status checking failure as the third priority because this issue might be caused by | ||
// a temporary glitch, for example, getting target information by using the Kubernetes API fails | ||
// after the pause operation. | ||
if (statusCheckFailedException != null) { | ||
if (pauserException == null) { | ||
pauserException = statusCheckFailedException; | ||
} else { | ||
pauserException.addSuppressed(statusCheckFailedException); | ||
} | ||
} | ||
|
||
// Treat the status unmatched issue as the third priority because this issue might be caused by | ||
// temporary glitch, for example, pod restarts. | ||
if (statusUnmatchedException != null) { | ||
if (pauserException == null) { | ||
pauserException = statusUnmatchedException; | ||
} else { | ||
pauserException.addSuppressed(statusUnmatchedException); | ||
} | ||
} | ||
|
||
return pauserException; | ||
} | ||
} |
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,11 @@ | ||
package com.scalar.admin.kubernetes; | ||
|
||
public class StatusCheckFailedException extends PauserException { | ||
public StatusCheckFailedException(String message) { | ||
super(message); | ||
} | ||
|
||
public StatusCheckFailedException(String message, Throwable cause) { | ||
super(message, cause); | ||
} | ||
} |
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,11 @@ | ||
package com.scalar.admin.kubernetes; | ||
|
||
public class StatusUnmatchedException extends PauserException { | ||
public StatusUnmatchedException(String message) { | ||
super(message); | ||
} | ||
|
||
public StatusUnmatchedException(String message, Throwable cause) { | ||
super(message, cause); | ||
} | ||
} |
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,11 @@ | ||
package com.scalar.admin.kubernetes; | ||
|
||
public class UnpauseFailedException extends PauserException { | ||
public UnpauseFailedException(String message) { | ||
super(message); | ||
} | ||
|
||
public UnpauseFailedException(String message, Throwable cause) { | ||
super(message, cause); | ||
} | ||
} |
Uh oh!
There was an error while loading. Please reload this page.