Issues fixed in MarkLogic Server 10.0-5

Unable to leave cluster with encryption enabled

Host cannot leave a cluster if config encryption is enabled because we leave encrypted config files on the host. One config file, server.xml, stores the license key of the host. If the customer doesn't not remember their license key, they will need to reach out to marklogic support to get their license key again.

Ordering geospatial point indices within cluster gives returns an error

XDMP-CANTDECODELOCATORORDERERKIND issued if the query issued on a cluster but works in a single node

Rebalancer does not resolve a deadlock

Under some rare circumstances, the rebalancer is unable to complete operations when it encounters an XDMP-DEADLOCK. Under normal circumstances, MarkLogic can attempt to resolve these - however under this scenario, it does not resolve the deadlock.

COMMIT on Cursor Close

In some circumstances, the ODBC driver does not issue a commit when a cursor is closed. We now ensure a commit is issued after a cursor is closed.

cts:search only performed on the first item in map

Given a cts:search query where a map is provided to a function that searches the items within the map, cts:search is only performed on the first item in the map with the correct language setting.

Performance slow down when exporting large amounts of numeric data via ODBC

A performance discrepancy has been observed where large numbers of records (> 100M), each containing 100+ columns are being downloaded over ODBC. To resolve this, the ODBC type modifier protocol has been implemented to send size information to the driver.

cts:stem only stems on the first item in the sequence in a nested eval

cts:stem() only stems on the first item in the sequence in a nested eval with the correct language setting.

Incorrect search results for cts:near-query that includes cts:not-query

Sometimes a cts:search misses out certain document matches when it includes a cts:near-query with a cts:not-query

Export of query deserialized by cts.query()

An Optic builder can specify a constraining cts.query in a where operation. Previously, if the cts.query was parsed by the SJS cts.query() function, an export() operation threw an error. Now, the export() operation succeeds, and the exported query plan can be reimported later and executed.

Inconsistent REST treatment of partition-number value

An invalid empty value for partition-number is exported by a REST GET; POST of an empty value succeeds but returns an error

Segmentation fault in Journal::writeJournal

Under some rare circumstances, an occasional segmentation fault (in Journal::writeJournal) can be observed, which will cause the MarkLogic process to restart.

GET "/manage/v2/protected-collections" Not Found, SEC-ROLEDNE

In some circumstances, an HTTP GET to "/manage/v2/protected-collections" incorrectly returns 404 Not Found, with message "SEC-ROLEDNE: (err:FOER0000) Role does not exist:..."

ODBC access log output lacking

ODBC access log doesn't have a log entry for errors thrown

False negatives for field-value-query

When stemming is enabled at the database level and disabled for a named field, field-value-query returns false negatives.

XCC/J Add Database (Contentbase) as a header

Database information is now added to the HTTP header in XCC.

Regression with null node permissions.

There is a regression updating a NullNode using xdmp.nodeReplace() resulted in "Permission denied: node-update", breaking existing application code.

Regression with cts:element-value-query for a search with a trailing wildcard

Under some circumstances a cts:element-value-query may return no results after upgrading from MarkLogic 9 to 10 due to the deprecation of an earlier format compatibility in MarkLogic 10.

Nitro EC2 instances with SSD instance store storage may cause Managed Cluster failure

Some Nitro based EC2 instance types that we support like r5d, r5ad, i3.metal, m5ad,m5d have SSD instance stores physically attached to them. When a Managed Cluster is start, it mounts the SSD instance store instead of the provisioned EBS volume, to /var/opt/MarkLogic. When the instance reboots the Marklogic comes up uninitialized because the SSD instance stores are ephemeral.

Deleting a database field with the Admin UI could cause data loss

There is an issue with drop-field-go.xqy in the Admin GUI: If a database field is created with the Admin API, it is possible that attempting to delete that field (or any field in that database) with the Admin UI will catastrophically delete the database configuration.

Replication database "Suspend" button is not greyed out once database is suspended

When Database Replication is configured between the primary database and DR, and when you use the suspend button in the database view in the Admin GUI for the primary database replication configuration, there is currently no indication that Database is suspended from within the Admin GUI.

Transient read-after-write failure under heavy load

On a heavily loaded cluster a transient read-after-write failure can occur following an XDQP restart.

Creating metadata fields with JSON does not work

Attempting to create a metadata field with a JSON payload will fail because the metadata property is lost.

Cluster scope support dump not working when a host of the cluster is offline

Having one (or more) non-bootstrap host offline prevents the customer from generating and submitting a support dump to Support services via telemetry.

Deleting path namespace and range path indexes fails

If you attempt to delete a path-namespace that was in use by a range-path-index, even if you were simultaneously deleting that range-path-index, the attempt fails

Database status page does not clearly state whether database replication is suspended or active

When Database Replication is configured between a primary database and the DR cluster, when you use the suspend button on the primary database using the database replication configuration, there is currently no clear indication that Database Replicatin is suspended between the two clusters.

AWS Managed Cluster mounts wrong EBS volume after termination/relaunch

If 2 EC2 instances in the same availability zone are terminated and relaunched at the same time, the hosts may swap primary EBS volumes.

Rebalancer with Segment or Legacy Assignment Policy

To correct a rare but significant bug in the Rebalancer, the Rebalancer may now take longer to complete when the database's forest assignment policy is set to 'legacy' or 'segment'.

Stemming performance

This is understood to be a MarkLogic Server performance regression that was introduced from the 9.0-11 release of MarkLogic 9 and which affects all subsequent releases. Under some circumstances, the ingestion of documents into a database with stemming enabled is significantly slower than would be expected. Ingestion for very large documents may time out as a result.

Segmentation fault in JournalArchive::writeJournal

Under some rare circumstances, segmentation faults have been observed in JournalArchive::writeJournal; this causes the MarkLogic Server process to restart.

JS-FROM-LIMITEXCEEDED error for /v1/rows request with a mapping function

The /v1/rows endpoint of the REST API can apply a mapping function to transform the result rows provided in the response. Previously, when applying the mapping function to more than 4096 rows, the request failed with a JS-FROM-LIMITEXCEEDED error. Now, the request can apply the mapping function to 262,144 rows in a single request. Regardless of whether the request applies a mapping function, if the query returns a large number of rows, the client should page over the result rows.

MacOS X SQL Driver not working with Tableau Correctly

Unable to connect using the DSN named "MarkLogicSQLDSN".

Issue with the description ofsample_text_summarizer_vocab.zip download on developer.marklogic.com

From the developer site, the sample_text_summarizer_vocab.zip is listed as being 1.1 MB, the download is actually 3.0 GB

MLCP logging should conain additional loging to correctly report total count of triples ingested

When loading RDF documents containing large numbers of RDF triples (e.g. an RDF file containing > 2M triples), MLCP currently does not log and aggregate the total count of triples ingested when it updates the percentage completed progress indicator.

Unable to issue /manage/v2?view=health request with role manage-user

Unable to issue /manage/v2?view=health request with role manage-user , error reports manage-admin privilege is required

Metadata dropped by mlcp import

When an mlcp import transform is specified, document metadata is dropped

Data in Tableau populates as today's date using a Live connection with queries containing TIMESTAMPADD()

From MarkLogic, the TIMESTAMPADD() function is not sending Tableau the correct ODBC type id's. This causes the data to be populated as today's date

Logging stops and cluster slows/hangs

With encryption at rest, logging stops and cluster slows/hangs at log rotate time

Segmentation fault on clusters during a rebalancing operation

Under some rare circumstances, a segfault can take place during an XDQP probeByHash operation. This can happen after an OS level hang and will cause the MarkLogic process to restart; subsequent rebalance operations will be retried after the restart.

Full outer join in the Optic API

Previously, the Optic API supported left outer joins and inner joins. Now, the Optic API also supports full outer joins

Database refragmenting gets stuck in an endless loop when a fragment's parent contains commented out element data

When an element configured as a fragment parent for a database has commented out elements in the document, any configuration changes to the database cause the reindex/refragment to get stuck in an endless loop.

Removing cluster host using endpoint may cause cluster to hang

Using the endpoint admin/v1/host-config to request that a host remove itself from the cluster may cause cluster to hang.

XDMP-CAST error for Entity Services conversion on JSON null nodes

Entity Services provides the es.instanceConverterGenerate() / es:instance-converter-generate() functions to generate a converter function for entity instances. Previously, if a JSON instance had a property with a JSON null value, the conversion function failed with an XDMP-CAST error. Now, the generated code skips JSON null nodes, preventing the casting error.

Admin GUI invalid coercion (XDMP-AS) where last-startup date is not available

Under some rare circumstances, after making a host level change, the following exception may be encountered: XDMP-AS: (err:XPTY0004) let $last-startup as xs:dateTime := xdmp:host-status($host)/hs:last-startup -- Invalid coercion: () as xs:dateTime In /wait-server-restart.xqy on line 177 At present, the $last-startup value is required but there may be circumstances where there is no value returned.

XDMP-BAD on values or value-match on empty sequence

On a clustered environment, when running against certain nodes, cts:element-value-match may incorrectly return XDMP-BAD when there is no match.

XDMP-KEYMISMATCH errors

Certain documents become unreadable with XDMP-KEYMISMATCH errors.

Unable to download binary files from Query Console using Explorer

Error when trying to download binary files from Qconsole documents explorer

Top X style SQL / SPARQL / Optic API queries are slow

Optic engine queries that use an order-by followed by a limit can be slow, as the entire sort is performed before any results can be returned.

Segmentation fault caused when parsing External Variables over an XDBCInvoke

Segmentation fault caused when parsing External Variables over an XDBCInvoke. When using Request.setNewVariables() on the client side, circumstances under which an empty value is bound to a variable as part of a sequence. The fix now handles this scenario.

Segmentation faults occurring while connected to Informatica Data Quality tool over ODBC

Under some rare circumstances, a segmentation fault can take place during the ODBCConnectionTask - this is usually the result of a hang during communication and these events usually happen in the background and have been observed whilst using Informatica Data Quality Tool.

MLCP does not retry failed commits

MLCP does not retry loading documents when a commit fails (mostly due to 502-504 errors) and the inserts in the current transaction will be lost

Cookie “SessionID” will be soon rejected because it has the “sameSite” attribute set to “none”

MarkLogic does not specify a "samesite" value for cookies which could break functionality on some Browsers because the default value "none" is used.

Whitespace stripped from parsed JSON text nodes.

If a JSON document contains a "language" or a "lang" property in an object, whitespace may be improperly stripped from text nodes when parsing.

Protected path not redacted with SQL

Under some circumstances, A SQL Query may expose data in a document that has been redacted using protected paths.

Protected Path On Property Hides All Entity Model Property Triples

Setting a protected path on one property hides all entity model property triples for other properties that are not protected. This bug also affects TDE templates that use wildcards in their context expressions. If one node returned by the context expression is protected, all other nodes matching the context expression will be protected as well.

Documentation for xdmp:sql incorrectly states that the built-in should not be used in production

In the document for xdmp:sql, the summary states that the function: Executes an ad hoc SQL query. This function is for testing your SQL views when data modeling; it is not intended to be used directly in applications. This is not true; xdmp:sql can be used in production.

Admin GUI: Last user containing the admin role should not be removed without prior confirmation

MarkLogic's Admin GUI (on port 8001) will currently allow you to remove a user with an admin role even if there are no other users with an admin role. This "final admin user" should not be removed without warning from MarkLogic Server and subsequent confirmation from the user.

AWS Managed Cluster may hang when a host is removed from the cluster

Running the command `mlcmd leave-cluster -terminate` may reset the hosts.xml on all remaining cluster hosts, causing the cluster to hang.

XDMP-INTERNAL exception thrown during SPARQL evaluation where externally bound variables are used within a group by clause

Under some circumstances, a SPARQL GROUP BY statement will fail with an XDMP-INTERNAL error. This happens whern the statement makes use of variables that are externally bound and which are resolved within the scope of a GROUP BY clause.

Rolling Upgrade to 10.0-4 - MLCP fails with a mixed versioned cluster

During rolling upgrade, if there is a mixed cluster that has 10.0-4 MarkLogic with lower versions (e.g. 10.0-3 with 10.0-4), running MLCP job with 10.0-4 MLCP may fail.

Cookie “SessionID” will be soon rejected because it has the “sameSite” attribute set to “none”

MarkLogic does not specify a "samesite" value for cookies which could break functionality on some Browsers because the default value "none" is used.

REST endpoint /manage/v2/databases/{id|name}/properties does not support creation of a field with name "language"

calling an HTTP PUT against /manage/v2/databases/{id|name}/properties with a payload containing the value "language" as a value of the JSON property (or XML element) "field-name" returns HTTP 400 status code with the error code (ADMIN-INVALIDCONFIG)

XDMP-ARGTYPE error thrown while trying to create a continuous line graph on Tableau

Under some rare circumstances an XDMP-BADCHAR will be thrown by MarkLogic when evaluating a TIMESTAMPADD along with a GROUP BY clause

OpenLDAP 2.4.50

OpenLDAP upgraded to 2.4.50

NVMe mount mapping changes after instance reboot

Nitro based instances have nvme device mappings for the volumes mounted to the instance. These mapping change when the system gets rebooted, This leads to cross mounting of different disks to incorrect mount points on reboot.

Creating an external security object LDAP and attaching it to an App server with changing port results in '500: Internal Server Error'

Creating an external security object LDAP and attaching it to an App server with changing port results in '500: Internal Server Error'

Incorrect search results on JSON documents with "lang" or "language" property

MarkLogic server may return incorrect search results after a reindex on JSON documents with the "lang" or "language" property set to a language other than the database language setting. Because this fix has backward compatibility issues, you must enable the "Reindex JSON language property” trace event to enable the fix; For additional context, contact MarkLogic Technical Support.

Correction in the datatype of RDF boolean literal

In the event where an RDF description of an item contains content expressed as an xs:boolean datatype, MarkLogic will use the wrong default datatype for the boolean value when storing the RDF content.

database-set-maintain-last-modified required privileges

Required privileges for database-set-maintain-last-modified (both admin API and RMA endpoint) are incorrect; Implementation now matches documentation at https://docs.marklogic.com/10.0/admin:database-set-maintain-last-modified .

"Top X" style SQL/SPARQL/Optic API query Performance

Performance improvement for "Top X" style SQL/SPARQL/Optic API queries, which previously took time proportional to the total data;

Dates and Datetimes showing up as today in Tableau reports

When connecting over ODBC using cursors, dates and datetime types get populated as today's date

SPARQL query performance regression

SPARQL performance can be slow in queries with != filters

Stack Trace for eval and invoke error

Missing stack trace in the error response for eval and invoke

SPARQL Query XDMP-MEMCANCELED

In some circumstances, SPARQL Query will be failing with XDMP-MEMCANCELED; For the cases identified, SPARQL query optimizations are improved with this release;

Weight parameters missing in the Search API

The weight parameter is supported in the cts:query() constructors for range and geospatial queries but not in the equivalent Search API queries.

Query Console taking long time to retrieve properties for large documents

Under some circumstances, where an XML document is large (in excess of 1 GB) and is stored alongside a small properties fragment, Query Console will hang when attempting to retrieve the properties.

Query Console keyboard lag

Query Console users experience noticeable lag times between keypress and when it renders on screen when there are a large number of queries contained in the workspace, or a large number of lines in the query tabs.