Known Issues for Event Stream Processor Server

By default, the RSA login uses the "SHA1withRSA" signature algorithm and the "MD5" digest algorithm. If you change the signature and digester methods in the cluster configuration, make the same changes to the SIGN_ALGORITHM and DIGEST_ALGORITHM environment variables.

For example, values for the SIGN_ALGORITHM environment variable are "SHA1withRSA" (default), "MD5withRSA", and "SHA1withDSA" (Java only). Possible values for the DIGEST_ALGORITHM environment variable are "MD5" (default) and "SHA1".

Avoid using retention with input windows that use a log store. While the compiler does not flag this as an error, the retention policy on input windows that use a log store may not work as expected after recovery.

In the case of the ESP Server crashing, if you have a join-based window using a log store, one source window using a log store, and another source window using a memory store and being derived from a window using a log store, the data in the join window is recovered up to the crash, but new records uploaded to the ESP Server do not get joined. When all the source windows use a log store, the same records that are uploaded to the ESP Server get joined.

If you are using an external function that returns a string, do not assign this string to the arena. Otherwise, an unrecoverable error occurs when the Server later tries to release the memory used by the returned string because the memory would have already been released by the Server when the record was processed.

To work around this issue, during a controlled shut down, close the Replication Server adapter before closing Event Stream Processor.

The ESP compiler views the string literal "\b" as a backspace character. Therefore, CCL programs that contain \b will not compile.

To work around this issue, use a double back slash ("\\b") for this string literal instead of a single back slash.

In Solaris AMD 11 allocating more than 4GB of memory may result in the server shutting down with an out of memory error. The error is related to loading the Java virtual machine, version 1.6.0_26 and later.

To work around this issue, specify a minimum java-max-heap project option with a value of 4096 to allocate up to 29GB of memory. To allocate up to 100GB of memory, set the java-max-heap project option to 32768.

Note that setting the java-max-heap option only sets the maximum heap size that Java can consume; it does not actually allocate memory.

To reduce memory consumption growth, when allocating a vector or a dictionary using 'new' that is subsequently being used as an argument to the getData() function, check for the vector already being allocated using isnull, as follows:

if (isnull(vector_var))
    vector_var = new vector(...)
...
...
getdata(vector_var, ...)