[jira] [Reopened] (OAK-7286) DocumentNodeStoreBranch handling of non-recoverable DocumentStoreExceptions
[
https://issues.apache.org/jira/browse/OAK-7286?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Marcel Reutegger reopened OAK-7286:
---
Need to re-open again. This change introduced a regression detected by
ReplicaSetResilienceIT. The test periodically shuts down the MongoDB primary
and checks if commits survive the failover. Previously the DocumentNodeStore
would retry the commit when there was a DocumentStoreException caused by the
primary shut down. The MongoDB driver then waits for the new primary to be
elected and the commit eventually goes through. Now the DocumentStoreException
is immediately propagated up the stack and the commit fails.
> DocumentNodeStoreBranch handling of non-recoverable DocumentStoreExceptions
> ---
>
> Key: OAK-7286
> URL: https://issues.apache.org/jira/browse/OAK-7286
> Project: Jackrabbit Oak
> Issue Type: Task
> Components: documentmk
>Reporter: Julian Reschke
>Assignee: Marcel Reutegger
>Priority: Major
> Fix For: 1.9.0, 1.10
>
> Attachments: OAK-7286.diff, OAK-7286.diff
>
>
> In {{DocumentNodeStoreBranch.merge()}}, any {{DocumentStoreException}} is
> mapped to a {{DocumentStoreException}} to a {{CommitFailedException}} of type
> "MERGE", which leads to the operation being retried, and a non-helpful
> exception being generated.
> The effect can be observed by enabling a test in {{ValidNamesTest}}:
> {noformat}
> --- oak-jcr/src/test/java/org/apache/jackrabbit/oak/jcr/ValidNamesTest.java
> (Revision 1825371)
> +++ oak-jcr/src/test/java/org/apache/jackrabbit/oak/jcr/ValidNamesTest.java
> (Arbeitskopie)
> @@ -300,7 +300,6 @@
> public void testUnpairedHighSurrogateEnd() {
> // see OAK-5506
>
> org.junit.Assume.assumeFalse(super.fixture.toString().toLowerCase().contains("segment"));
> -
> org.junit.Assume.assumeFalse(super.fixture.toString().toLowerCase().contains("rdb"));
> nameTest("foo" + SURROGATE_PAIR[0]);
> }
> @@ -336,6 +335,7 @@
> assertEquals("paths should be equal", p.getPath(), n.getPath());
> return p;
> } catch (RepositoryException ex) {
> +ex.printStackTrace();
> fail(ex.getMessage());
> return null;
> }
> {noformat}
> The underlying issue is that {{RDBDocumentStore}} is throwing a
> {{DocumentStoreException}} due to the invalid ID, and repeating the call will
> not help.
> We probably should have a way to dinstinguish between different types of
> problems.
> I hacked {{DocumentNodeStoreBranch}} like that:
> {noformat}
> ---
> oak-store-document/src/main/java/org/apache/jackrabbit/oak/plugins/document/DocumentNodeStoreBranch.java
> (Revision 1825371)
> +++
> oak-store-document/src/main/java/org/apache/jackrabbit/oak/plugins/document/DocumentNodeStoreBranch.java
> (Arbeitskopie)
> @@ -520,8 +520,12 @@
> } catch (ConflictException e) {
> throw e.asCommitFailedException();
> } catch(DocumentStoreException e) {
> -throw new CommitFailedException(MERGE, 1,
> -"Failed to merge changes to the underlying
> store", e);
> +if (e.getMessage().contains("Invalid ID")) {
> +throw new CommitFailedException(OAK, 123,
> +"Failed to store changes in the underlying
> store: " + e.getMessage(), e);
> +} else {
> +throw new CommitFailedException(MERGE, 1, "Failed to
> merge changes to the underlying store", e);
> +}
> } catch (Exception e) {
> throw new CommitFailedException(OAK, 1,
> "Failed to merge changes to the underlying
> store", e);
> {noformat}
> ...which causes the exception to surface immediately (see
> https://issues.apache.org/jira/secure/attachment/12912117/OAK-7286.diff).
> (cc [~mreutegg])
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Reopened] (OAK-7286) DocumentNodeStoreBranch handling of non-recoverable DocumentStoreExceptions
[
https://issues.apache.org/jira/browse/OAK-7286?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Marcel Reutegger reopened OAK-7286:
---
No, that shouldn't be the case. I think it should be wrapped into a
RepositoryException.
> DocumentNodeStoreBranch handling of non-recoverable DocumentStoreExceptions
> ---
>
> Key: OAK-7286
> URL: https://issues.apache.org/jira/browse/OAK-7286
> Project: Jackrabbit Oak
> Issue Type: Task
> Components: documentmk
>Reporter: Julian Reschke
>Assignee: Marcel Reutegger
>Priority: Major
> Fix For: 1.9.0, 1.10
>
> Attachments: OAK-7286.diff, OAK-7286.diff
>
>
> In {{DocumentNodeStoreBranch.merge()}}, any {{DocumentStoreException}} is
> mapped to a {{DocumentStoreException}} to a {{CommitFailedException}} of type
> "MERGE", which leads to the operation being retried, and a non-helpful
> exception being generated.
> The effect can be observed by enabling a test in {{ValidNamesTest}}:
> {noformat}
> --- oak-jcr/src/test/java/org/apache/jackrabbit/oak/jcr/ValidNamesTest.java
> (Revision 1825371)
> +++ oak-jcr/src/test/java/org/apache/jackrabbit/oak/jcr/ValidNamesTest.java
> (Arbeitskopie)
> @@ -300,7 +300,6 @@
> public void testUnpairedHighSurrogateEnd() {
> // see OAK-5506
>
> org.junit.Assume.assumeFalse(super.fixture.toString().toLowerCase().contains("segment"));
> -
> org.junit.Assume.assumeFalse(super.fixture.toString().toLowerCase().contains("rdb"));
> nameTest("foo" + SURROGATE_PAIR[0]);
> }
> @@ -336,6 +335,7 @@
> assertEquals("paths should be equal", p.getPath(), n.getPath());
> return p;
> } catch (RepositoryException ex) {
> +ex.printStackTrace();
> fail(ex.getMessage());
> return null;
> }
> {noformat}
> The underlying issue is that {{RDBDocumentStore}} is throwing a
> {{DocumentStoreException}} due to the invalid ID, and repeating the call will
> not help.
> We probably should have a way to dinstinguish between different types of
> problems.
> I hacked {{DocumentNodeStoreBranch}} like that:
> {noformat}
> ---
> oak-store-document/src/main/java/org/apache/jackrabbit/oak/plugins/document/DocumentNodeStoreBranch.java
> (Revision 1825371)
> +++
> oak-store-document/src/main/java/org/apache/jackrabbit/oak/plugins/document/DocumentNodeStoreBranch.java
> (Arbeitskopie)
> @@ -520,8 +520,12 @@
> } catch (ConflictException e) {
> throw e.asCommitFailedException();
> } catch(DocumentStoreException e) {
> -throw new CommitFailedException(MERGE, 1,
> -"Failed to merge changes to the underlying
> store", e);
> +if (e.getMessage().contains("Invalid ID")) {
> +throw new CommitFailedException(OAK, 123,
> +"Failed to store changes in the underlying
> store: " + e.getMessage(), e);
> +} else {
> +throw new CommitFailedException(MERGE, 1, "Failed to
> merge changes to the underlying store", e);
> +}
> } catch (Exception e) {
> throw new CommitFailedException(OAK, 1,
> "Failed to merge changes to the underlying
> store", e);
> {noformat}
> ...which causes the exception to surface immediately (see
> https://issues.apache.org/jira/secure/attachment/12912117/OAK-7286.diff).
> (cc [~mreutegg])
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
