[
https://issues.apache.org/jira/browse/CASSANDRA-20389?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17931910#comment-17931910
]
Sam Tunnicliffe commented on CASSANDRA-20389:
---------------------------------------------
{quote}Fixing this table name validation bug depends if the validation is
called only for new tables, or also for existing tables, when the schema is
shared with another node or replica
{quote}
The validation will be enforced whenever a new table, index or view is created
as well as when a table is altered. At these times, every node will perform the
validation check independently as the operation is replicated and each peer
executes it. Additionally, as the operation has been committed to the metadata
log, it will get replayed at startup which is what's responsible for stacktrace
[~maxwellguo] posted. A similar issue does also exist in earlier versions
though, because the table name is written to the {{system_schema.tables}} table
and reconstructed at startup whereby the call to
{{ColumnFamilyStore::scrubDataDirectories}} also errors out, causing a failure.
This should actually simplify things as it should be safe to assume that there
aren't clusters out there with these abnormally long table names which will be
broken if we fix the validation.
{quote}bq. There is a bug in
[TableMetadata|https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/schema/TableMetadata.java#L552]
which calls a name validation function with expectation of checking the
length. However, the correct function to use is
[isValidName|https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/schema/SchemaConstants.java#L83],
which validates also length, instead of
[isNameValid|https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/schema/TableMetadata.java#L552],
which, I guess, shouldn't be present.
{quote}
This looks like something was overlooked either in CASSANDRA-9456 or
CASSANDRA-16956. The latter added a comment which says
{code:java}
* Before CASSANDRA-16956, we used to care about not having the entire path
longer than 255 characters because of
* Windows support but this limit is by implementing CASSANDRA-16956 not in
effect anymore. {code}
It's probably correct that the length of the entire path is not an issue
outside of Windows, but there are still limits (clearly) on the length of an
individual file name (255 on most UNIX file systems IIRC). The constraint in
{{isValidName}} is to limit filenames to a max length of 48 chars, so I don't
think it's going to be safe to simply switch to using that as there may be
existing tables which don't cause the 255 limit to be breached, but would fail
with a 48 char limit. So I think we need another option to ensure that any
table name does not cause the FS limit of 255 to be exceeded, which must
include the table id as that is appended to the table name to form the name of
the data directory.
In future, we could move to naming the data dirs solely using the table id as
these are no longer re-used when a table is dropped and subsequently recreated.
This would make the dir names fixed length, but it would remove the ability of
operators to easily identify the data dir for a given table.
> Create table fails on long table names
> --------------------------------------
>
> Key: CASSANDRA-20389
> URL: https://issues.apache.org/jira/browse/CASSANDRA-20389
> Project: Apache Cassandra
> Issue Type: Bug
> Reporter: Ruslan Fomkin
> Priority: Normal
>
> If CREATE TABLE is called with a long table name, the statement fails due to
> file name too long error.
> E.g, such test:
> {code:java}
> @Test
> public void testCreatingTableWithLongName() throws Throwable
> {
> String keyspace = "\"38373639353166362d3566313\"";
> String table =
> "test_create_k8yq1r75bpzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzaaaaaaaaaaaaaaaa";
> execute(String.format("CREATE KEYSPACE %s with replication = " +
> "{ 'class' : 'SimpleStrategy', 'replication_factor'
> : 1 }",
> keyspace));
> createTableMayThrow(String.format("CREATE TABLE %s.%s (" +
> "key int PRIMARY KEY," +
> "val int)", keyspace, table));
> } {code}
> results in:
> {code:java}
> java.lang.RuntimeException: Failed to list files in
> build/test/cassandra/data/38373639353166362d3566313/test_create_k8yq1r75bpzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzaaaaaaaaaaaaaaaa-1b255f4def2540a60000000000000007
> at
> org.apache.cassandra.db.lifecycle.LogAwareFileLister.list(LogAwareFileLister.java:77)
> at
> org.apache.cassandra.db.lifecycle.LifecycleTransaction.getFiles(LifecycleTransaction.java:636)
> at
> org.apache.cassandra.db.Directories$SSTableLister.filter(Directories.java:1139)
> at
> org.apache.cassandra.db.Directories$SSTableLister.list(Directories.java:1086)
> at
> org.apache.cassandra.db.Directories$SSTableLister.list(Directories.java:1072)
> at
> org.apache.cassandra.db.Directories.lambda$getUIDGenerator$5(Directories.java:1297)
> at
> java.base/java.util.stream.AbstractPipeline.sourceSpliterator(AbstractPipeline.java:405)
> at
> java.base/java.util.stream.AbstractPipeline.evaluate(AbstractPipeline.java:234)
> at java.base/java.util.stream.IntPipeline.reduce(IntPipeline.java:496)
> at java.base/java.util.stream.IntPipeline.max(IntPipeline.java:459)
> at
> org.apache.cassandra.io.sstable.SequenceBasedSSTableId$Builder.generator(SequenceBasedSSTableId.java:102)
> at
> org.apache.cassandra.db.Directories.getUIDGenerator(Directories.java:1302)
> at
> org.apache.cassandra.db.ColumnFamilyStore.createColumnFamilyStore(ColumnFamilyStore.java:768)
> at
> org.apache.cassandra.db.ColumnFamilyStore.createColumnFamilyStore(ColumnFamilyStore.java:756)
> at
> org.apache.cassandra.db.ColumnFamilyStore.createColumnFamilyStore(ColumnFamilyStore.java:747)
> at org.apache.cassandra.db.Keyspace.initCf(Keyspace.java:379)
> at
> org.apache.cassandra.schema.DistributedSchema.createTable(DistributedSchema.java:341)
> at
> org.apache.cassandra.schema.DistributedSchema.lambda$initializeKeyspaceInstances$6(DistributedSchema.java:213)
> at java.base/java.lang.Iterable.forEach(Iterable.java:75)
> at
> org.apache.cassandra.schema.DistributedSchema.lambda$initializeKeyspaceInstances$10(DistributedSchema.java:213)
> at com.google.common.collect.ImmutableList.forEach(ImmutableList.java:422)
> at
> org.apache.cassandra.schema.DistributedSchema.initializeKeyspaceInstances(DistributedSchema.java:199)
> at
> org.apache.cassandra.tcm.listeners.SchemaListener.notifyInternal(SchemaListener.java:50)
> at
> org.apache.cassandra.tcm.listeners.SchemaListener.notifyPreCommit(SchemaListener.java:43)
> at
> org.apache.cassandra.tcm.log.LocalLog.notifyPreCommit(LocalLog.java:626)
> at
> org.apache.cassandra.tcm.log.LocalLog.processPendingInternal(LocalLog.java:527)
> at
> org.apache.cassandra.tcm.log.LocalLog$Async$AsyncRunnable.run(LocalLog.java:819)
> at
> org.apache.cassandra.concurrent.InfiniteLoopExecutor.loop(InfiniteLoopExecutor.java:121)
> at
> io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
> at java.base/java.lang.Thread.run(Thread.java:829)
> Caused by: java.nio.file.FileSystemException:
> build/test/cassandra/data/38373639353166362d3566313/test_create_k8yq1r75bpzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzaaaaaaaaaaaaaaaa-1b255f4def2540a60000000000000007:
> File name too long
> at
> java.base/sun.nio.fs.UnixException.translateToIOException(UnixException.java:100)
> at
> java.base/sun.nio.fs.UnixException.rethrowAsIOException(UnixException.java:111)
> at
> java.base/sun.nio.fs.UnixException.rethrowAsIOException(UnixException.java:116)
> at
> java.base/sun.nio.fs.UnixFileSystemProvider.newDirectoryStream(UnixFileSystemProvider.java:412)
> at java.base/java.nio.file.Files.newDirectoryStream(Files.java:472)
> at
> org.apache.cassandra.db.lifecycle.LogAwareFileLister.innerList(LogAwareFileLister.java:83)
> at
> org.apache.cassandra.db.lifecycle.LogAwareFileLister.list(LogAwareFileLister.java:73)
> {code}
> There is a bug in
> [TableMetadata|https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/schema/TableMetadata.java#L552]
> which calls a name validation function with expectation of checking the
> length. However, the correct function to use is
> [isValidName|https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/schema/SchemaConstants.java#L83],
> which validates also length, instead of
> [isNameValid|https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/schema/TableMetadata.java#L552],
> which, I guess, shouldn't be present.
> Fixing this table name validation bug depends if the validation is called
> only for new tables, or also for existing tables, when the schema is shared
> with another node or replica.
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
