Warn about metadata and heap size ?

[Tom Zeller]

> On Sep 30, 2015, at 12:02 PM, Cantor, Scott <cantor.2 at osu.edu> wrote:
> 
> On 9/30/15, 11:42 AM, "dev on behalf of Rod Widdowson" <dev-bounces at shibboleth.net on behalf of rdw at steadingsoftware.com> wrote:
> 
>>> Just wondering if in v3 we want the metadata resolver to warn and possibly
>>> not reload if a metadata file size is larger than some value. For example, -
>>> Xmx768m does not seem quite enough for the ~20m UK federation
>>> metadata. I guess I would rather see a log message at warn that says “raise
>>> your heap size” than “java.lang.OutOfMemoryError”, unless of course the
>>> latter is recoverable.
>> 
>> My instinct is that this is being overly cute.  It’s the misanthropist in me, but I cannot help but believe that the people who might benefit from this wouldn't see the warning, and the ones who would see the warning wouldn't need it.
> 
> I had basically concluded the most we might do is just warn on smaller heaps at startup, but I also realize that isn't going to really matter much either. I just thought it would be a way of shutting down complaints that we changed the guidance on this without warning.
> 
> I agree that changing Out of Memory to "increase your heap size" is kind of silly. As long as we're suitably loud about what the failure was (failure to reload the metadata).
> 
> I wish there was a CRIT category like the SP has.

Even though I am aware of the issue, I was still surprised that with -Xmx768m and a single 20m metadata file, adding a tiny metadata file and reloading the metadata resolver would result in essentially a crash and the need to restart the app. Log message aside, I guess what I am really asking is : should we avoid reloading the resolver if we think it will “crash” the IdP ?