When the keyStore contains trustedCertEntry entries alongside the PrivateKeyEntry and the PrivateKeyEntry is not the 1st entry in the keyStore, Jmeter 2.7's JmeterKeystore.load cannot find the key. It throws an exception which accompanying message reads "No key(s) found".

The reason lies in the implementation of the load method. Its skeleton when scanning the aliases is:

         if (null != is){ // is is the InputStream
            PrivateKey _key = null;
            int index = 0;
            Enumeration<String> aliases = store.aliases();
            while (aliases.hasMoreElements()) {
                String alias = aliases.nextElement();
                if (store.isKeyEntry(alias)) {
                    if ((index >= startIndex && index <= endIndex)) {
                        _key = (PrivateKey) store.getKey(alias,...);
                        if (null == _key) {
                            throw new Exception(...);
                        }
                        ...
                        v_names.add(alias);
                        ...
                    }
                }
                index++;
            }

            if (null == _key) { // Defect: source of problem
               throw new Exception("No key(s) found");
            }
        }
        int v_size = v_names.size();
        ...

So:
- The location test of _key itself would be a problem because _key would always be the last entry read in the keyStore, which might not be the private key.
- But the fact that startIndex and endIndex are 0 (default initialisation values of implicitly initialised arguments, see SSLManager.java) implies that the private key would be found only if it was the 1st entry in the keystore. I didn't find any hint that this is a JSSE requirement.

Assuming that only 1 key can be loaded (another source file states that no provision has been made to allow the user to specify one key amongst many) I think that:
- the "if ((index >= startIndex && index <= endIndex))" condition gets in the way;
- the validation that a key does exist in the keystore would be better done by asserting that "v_size != 0".

Note that JMeter 2.4 was loading the keystore along the lines I'm suggesting, which is no surprise since I located the problem by investigating how JMeter 2.4 was getting it right when 2.7 was failing (on the same keystore).