Issues fixed in MarkLogic Server 10.0-2

SPARQL CONSTRUCT intermittently returns a different count when ORDER BY is used

A SPARQL CONSTRUCT statement can occasionally return an extra triple when it returns results that are sorted using an ORDER BY

SQL Optimizer is not optimizing OR operator

Under some conditions, an SQL SELECT with nested AND/OR statements are optimized for a probable push down to the index filtering.

Installed Signed Certificate Details Not visibile in Certificate Template GUI

When Signed Certificate is installed through Admin GUI (or through Query Console ), User/Customer cannot see details on installed Certificate, including Certificate Serial Number, Certificate Issuer, Certificate CN field. Only data visible on Installed Cert is how many days Certificate is valid for in Cert Template status.

More stable cluster heartbeat

Now using more concurrency when broadcasting heartbeats to other hosts. This make heartbeats more reliably delivered from healthy hosts to healthy hosts in the cluster. And prevents unhealthy hosts in the cluster from causing other hosts to be unhealthy. Resulting in fewer XDQP restarts.

CNTK function constructor cannot take trained models as documents

The CNTK function constructor errors when passed a neural network model as a document.

The REST Management API logs endpoint ignores the host parameter

Irrespective of the host specified, the logs endpoint only lists the log files on the current host. This defect has been corrected.

universally quantified not equals operator in triple index

A universally quantified not equals operator is now supported in the triple index

Segmentation fault in xdmp::ParseNode::rethrow

Unable to deploy DataHub project containing empty collations

Deploying DataHub project causes a segmentation fault in the implementation of xdmp:collation-canonical-uri

amp roles not updated when inherited role updated.

amp roles are not updated when inherited role is updated.

CPF Collection

CPF collections incorrectly references collation

Roles & MarkLogic Security database rendering performance

Rendering security database resources (via ReST API or Admin GUI) can take a long time to load if there are a very large number (thousands) of roles configured

Admin UI: Unexpected ADMIN-BADCERTTEMPLATE:

If an old template ID exists in configuration, the Admin UI will return a 500 Internal Server Error - ADMIN-BADCERTTEMPLATE - when changing any configuration item on any Application Server. This may occur after restoring a Security database.

Segmentation Fault in geospatial region query with null or empty geospatial path reference

MarkLogic Server crashes with a segmentation fault when a null geospatial path reference in JavaScript or an empty sequence for geospatial path reference in XQuery is passed to geospatial region query

CNTK Learner APIs Throw XDMP-INTERNAL instead of XDMP-ARG for Bad Schedule Argument

Learner constructors called with minibatch-size-schedules that are expecting learning-rate-schedules or similar do not mention arguments as reason in exception message.

Restore from S3 fails with SVC-S3SOCERR: SVC-SOCRECV: Connection reset by peer

Network connections sometimes get unexpectedly reset by S3 while MarkLogic is reading from it. If this happens while copying a file from S3, MarkLogic treats the connection reset as a file copy error. If the file is being copied as part of a forest restore, the forest restore fails. If the forest restore is part of a database restore, the database restore fails. MarkLogic now treats S3 file copy connection resets as retryable errors. It retries reading the file a few times before giving up. Persistent S3 problems will still cause restores to fail. But isolated transient S3 connection resets during file reads are recoverable.

Query console auto complete/suggestion doesn't work for any of the new cntk functions

cntk functions newly added in 10.0-1 not working in Query console auto complete/suggestion

Missing Memory Details in Monitoring History

In monitoring history > memory detail, Data File Memory, Forest Memory, Unclosed Stand Memory, Cache Memory, Registered Query Memory, and Join Memory are not shown.

Upgrade OpenSSL to 1.0.2s

MarkLogic 10.0.2 OpenSSL libraries have been updated to version 1.0.2s, providing additional encryption key algorithms including Elliptic-curve Diffie–Hellman (ECDH) and Elliptic Curve Digital Signature Algorithm (ECDSA)

Azure Storage Account specified in azure file paths

MarkLogic now accepts Azure Storage account names in azure file paths. Storage accounts are specified in azure file paths with an '@' following the container. Example: azure://mycontainer@myaccount/mydirectory/myfile Azure credentials in the security database are used if the storage account matches. Otherwise Azure IAM credentials are required.

spaces in implicit value queries in QBE

By default, QBE executes a value query for criteria that is not explicitly a value, word, or range query. Previously, QBE normalized spaces within an implicit value query, which can cause criteria with multiple contiguous spaces to fail to match. Now, QBE preserves spaces for implicit value queries.

local disk failover forest unavailable after master forest restored.

local disk failover forest is not available after the master forest is restored.

xdmp:encoding-language-detect throws XDMP-UTF8SEQ

xdmp:encoding-language-detect throws XDMP-UTF8SEQ when certain characters are present:, for example characters such as U+0086 that are valid as escaped sequences in XML 1.1.

Hash collision in distinct operator

SPARQL hash collision in distinct operator resulted in SQL distinct not returning all the unique values

Segmentation fault in XDQPServerSession::lockStatus

Segmentation fault occurs, with logs indicating that XDQPServerSession::lockStatus is the cause of the fault.

Azure Key Vault for Encryption at Rest

Native support of Azure Key Vault for Encryption at Rest does not exist

CNTK Functions allow invalid parameters in cntk:parameter

Calling a CNTK builtin with nonexistent parameters or misspelled optional parameters (such as iniit or map-rank-large) do not throw exceptions.

Bulk write with a transform in the REST API fails on Windows

Performing a Bulk write with a transform via the REST API in any supported Windows OS fails and results in a similar error(below) within the ErrorLog.txt file: ========= Info: Status 500: SVC-FILSTAT: cts:tokenize("attachment; filename="/IBC/container/T40819504.xml"; c...", "http://marklogic.com/collation/") -- File status error: GetFileAttributes 'E:\Program Files\MarkLogic\Data\http://marklogic.com/collation/-simple.unified-cd': The filename, directory name, or volume label syntax is incorrect. =========

searchable expressions deprecated in the Search, REST, and Java APIs

Previously, the Search API supported a searchable-expression within query options. As a result, the REST API and Java API layers above the Search API supported searchable-expression as well. The searchable expression is an XPath expression equivalent to the first argument of cts:search(). The searchable expression is executed with a variable mix of queries on indexes and filtering on results. The searchable-expression is problematic for performance reasons and, within the Search API, for security reasons. In particular, a searchable-expression from an untrusted source must be sanitized with search:check-options() or with the validate-options property of the REST API (which is true by default). Now, the searchable-expression is deprecated. Applications can continue to use searchable-expression during the deprecation period by assigning the eval-search-string privilege to one of the user's roles or by using a searchable-expression that validates against the cts:valid-extract-path() function. Applications should move to queries and (where needed) post-processing to filter on results.

Obsolete stands not cleared

In certain circmstances, stands no longer in use are not cleared or are only cleared after an extended periond of time.

Azure Key Vault for Encryption at Rest

Native support of Azure Key Vault for Encryption at Rest does not exist

geospatial expression functions missing in Optic

Optic supports builtin expression functions for manipulating values while retrieving and processing rows. Previously, geospatial expression functions were not available for manipulating values.

Support dump includes 401 unauthorized for multiple hosts in a cluster

When a support dump is created for multiple hosts in a cluster, it shows 401 unauthorized and is missing support information that is need to be collected.

SPARQL slowdown on upgrade from 8

Some SPARQL queries with bind variables slow down on upgrade from version 8

Can't change values in BackupTag.txt file when encryption enabled

Can't change values in BackupTag.txt file when encryption enabled

Segmentation fault in contentbuf::error accessing s3

MarkLogic may segfault processing an error condition while accessing s3.

Segmentation fault in QueryPathVector::setTimeLimit

Failing forests back after host maintenance causes an unacceptable database outage of over a minute

Failing forests back can - in some cases - lead to a database outage of 1 to 2 minutes and this means they need to make the case for a significantly larger maintenance window to account for this.

JSON search response returning a query having paths with namespaces

The /v1/search endpoint of the REST API supports getting a search response as JSON. If the query options sets the return-query flag to true, the search response serializes the query as JSON. Starting in 9.0-5, such requests threw a XDMP-UNBPRFX error. Now the request completes successfully.

Entity Services generates poor templates when nested entities are involved

The templates generated by Entity Services do not work properly when nested entities are involved and need to be edited by hand to make them work.

Server doesn't support sending app logs to Ops Director

Prior to 10.0-2, MarkLogic Server only supports sending system error logs to Ops Director. App Server error logs and Task Server error logs are not supported.

mustache library removed

The mustache library is unsupported and no longer used by MarkLogic Server. it is being removed from MarkLogic Server deliverables.

Database Name Support For Amps

Amps created on a cluster, do not work on a replicated cluster.

sparql query performance

sparql query performance improvements

Customized "concurrent request limit" for external LDAP authentication fails.

A Customized "concurrent request limit" value(set to 5 or lower) for external LDAP authentication returns "403 Forbidden" error. Internal LDAP authentication remains successful.

Optic API op.groupConcat does not take separator into account

Optic API op.groupConcat does not take separator character "|" into account when using SJS

Entity Services mapping XSLT-UNBPRFX error.

es.mappingPut() on a source mapping document generates XSLT-UNBPRFX error -- Prefix xs has no namespace binding.

CNTK Schedule Type Error Message breaks HTTP Header Parsing

Using the wrong type of CNTK parameter schedule in the additional learning options parameter of a learner constructor function breaks the HTTP Header message for the exception message.

Cannot configure multiple, external KMS servers with the REST Management API

The new functions for supporting HA on external KMS servers were not supported by the REST Management API. This has been fixed.

Search API requests without use of server fields

Previously, the Search API controlled diagnostic output by checking for a debug flag in a server field. Under rare conditions, when invoked repeatedly in a loop, the Search API used to throw an xdmp-get-server-field privilege error regardless of whether the debug flag was set or not. Now, the Search API uses a trace event to control diagnostic output, so the Search API never needs the xdmp-get-server-field privilege.

Linux cached pages more aggressively freed after reads

On Linux, to assist kernel memory management, MarkLogic now makes additional kernel calls advising of file pages it likely doesn't need. MarkLogic now advises it specifically doesn't need large data file pages immediately after it completes random-access reads of them. The intent is for the kernel to use this advice to more efficiently free memory from the page cache.

Document not identified in cts:reverse-query

Under some circumstances a cts:reverse-query failed to identify a document that described a cts:not-query immediately after the document had been inserted.

Segmentation fault retrying because of XDMP-UNREGISTERED

Server restarts showing a segmentation fault in DatabaseValuesFuture::isConcurrent while retrying because of XDMP-UNREGISTERED.

Database replication setup fails when using IE

When using IE to setup database replication, the Admin UI may generates a GET request that is too long for IE to support, causing the request to time out with few descriptive errors.

Selenium test needed to test this bug

The last update requires a selenium test for the qconsole code to be created and run in b10_0 and trunk to ensure that this issue doesn't happen again.

Error configuring XCC as JNDI resource on WildFly

Currently, customers are not able to configure XCC correctly as JNDI resource on WildFly.

Reindexer information in the Admin GUI incorrectly reports that reindexing has completed

In some rare cases, after making a modification to a TDE view can cause the Admin GUI to falsely report that reindexing has run to completion, whereas an attempt to project that view yields an "SQL-TABLEREINDEXING: Table is being reindexed" exception message. After this fix, the Admin GUI will correctly report the current state of the reindexer.

Admin UI cluster shutdown/restart buttons only applies to one group.

On the "Cluster Status" page of the Admin GUI, the "shutdown" button only shuts down hosts in the current group. Similarly the "restart" button only restarts hosts in the current group.

CNTK Crash

CNTK segmentation fault when the wrong model format is used,

Monitoring history refresh button removes the time span/Period selection

Observed Behavior: When timespan/period is refreshed in monitoring history, selection defaults to Raw irrespective of time span selection before refresh. Expected Behavior: Time span selection should not default to Raw when monitoring history is refreshed and it should preserve time span selection.

Entity Services creates inappropriate default collation

Generated database properties and search options created by Entity Services provide an inappropriate default collation. It is always the English collation. It should be the root collation for strings and the codepoint collation for anyURI data.

vulnerability in searchable-expression

The Search API provides the searchable-expression element in query options for setting the first argument to cts:search() The Search API evaluates this string, thus creating an opportunity for injection attacks though mitigated by the privileges of the roles of the user executing the query. The search:check-options() function now validates that the text of the searchable expression is a single expression and thus guards against injection attacks. Adopters accepting query options from external sources should either use search:check-options() or, in the REST API, leave the validate-options property with its default value of true.

Query Console: Incorrect stack frames in JavaScript

In Query Console, missing or duplicate error stack frames are returned for certain JavaScript errors.

xdmp:keystore-validate-export( ) doesn't work

xdmp:keystore-validate-export( ) doesn't work

Aggregate memory parameters incorrect in monitoring history

In the Monitoring History page, some hourly and daily aggregate memory parameters are incorrect. Forest Memory, Unclosed Stand Memory, Cache Memory, Registered Query Memory are all affected.

Journal archive directory is not purged of old versions by scheduled backups.

Alternate journal archive directories are not automatically purged of old sub-directories in scheduled backups.