Correlator

There is a known issue with running Apama on systems configured to use libnss_ldap in-process. This applies to any system authentication method using SSLwhich does not use an external daemon. This is due to the custom OpenSSL shipped with Apama. The work around is either to disable SSL for your LDAP requests or, preferably, configure the system with an external authentication provider such as SSSD or nslcd, rather than libnss_ldap directly.

Sending in large (8Mb) managed object uses disproportionately large amounts of memory ( > 1Gb).

The ordering of entries in containers within EPL and within connectivity plug-ins is undefined in some cases. A change in this release to reduce memory usage had the side effect of changing some of this ordering. This is particularly likely to be reflected in the order of entries in maps when printed by some connectivity plug-ins like the diagnostic codec or the JSON format produced by the JSON plug-in. It could also affect the order that map entries are iterated over in your own plug-ins.

EPL on completed InnerEvent() crashes the correlator.

A correlator crash has been fixed. The crash occurred when a monitor was deleted which contained both an event type defined within its scope (an “internal event”) and an on completed listener for such events.

Correlator should allow starting during grace period after license expiry.

The correlator now starts with an expired license in its grace period, that is, 7 days post expiration date.

Incorrect type for runtime object of any.getGenericAction().

Calling getGenericAction on an any value that is an action, a value of type action<sequence<any> > returns any is returned. However, after assigning the return to an any, a call to getTypeName() would return the original action's type which would match the original signature, not action<sequence<any> > returns any, for example, would be a type mismatch. The correct type is now returned by getTypeName().

GenericComponentManagementAdapter::getHostname deallocates its return value.

An issue leading to corrupted data in the Prometheus hostname metric has been fixed.

Combination of big events and channel prefixes can cause a crash in engine_send, correlator (EventStreamReader).

Parsing an event (.evt) file with single lines in excess of 256kB combined with lines using a channel name before the event could result in a crash. This issue is resolved.

Correlator crash handler should log EPL monitor name etc. if possible.

In case of a correlator crash, after the correlator stacktrace, the EPL stacktrace is now printed as well.

New soft-reserved keywords have been added to the Event Processing Language (EPL). The await keyword is now a soft-reserved keyword.

This may become a language feature in the future, so it is recommended to avoid using it to avoid confusion.