On 09/08/2014 11:42 AM, Ludovic Rousseau wrote:
2014-09-08 11:19 GMT+02:00 Umberto Rustichelli <[email protected]>:
OK, tests for millions of signatures were mostly positive for you (you all
in the list up to now) and also for us (things always break in production,
don't you know?).
The main difference is that both my production environment and test
environment work with 14 cards, with 2 USB hubs, 7 cards per hub (not one
card) and I never re-login/reconnect, just go on signing exploiting the
existing PKCS#11 session.
The only difference between my test and production environment were the
cards (the customer's cards are a bit more recent).
I think that we're going nowhere at the moment (my fault, given that I
cannot provide much debug data), anyway I'll keep listening if anyone can
help... I suppose that if we broaden the suspects pool to include a USB
glitch (excessive power drain that hampers USB communication) we are going
too far...
It may be a temporary hardware issue.
But in this case you should not have "apparently every time we restart
the cards the time-to-failure gets shorter"
Do you have numbers about the "gets shorter"? It is 10%, 50% or 90% shorter?
~80% would be a good guess. The customer is lazy and doesn't provide the
count of requested signatures, so the estimate is really bad.
For sure, they had 14 cards doing fine for one month. After the shutdown
and restart, we had the first card failing after two days, more followed
in five days.
Do you have the problem with different readers from different manufacturers?
Cannot test other readers because they wouldn't fit in the appliance for
the way it is built (we use extra-slim readers, seven cradled in a 2.5"
bay).
To be honest, we never experienced this sort of issues with this
specific test, but -being many years we play with smart cards- I recall
two situations that we can combine and obtain a scenario similar to the
current one, but it's not convincing to me.
Recently, we had a card that failed after some n * 10K signatures and
needed to be (physically) disconnected to work again.
A couple of years ago, instead, we had a card that failed miserably
(among 98 cards connected) and took down the whole lot in a few seconds;
I suppose the card was drawing too much energy and the USB chain, from
the leaf to the root hub, shut down. The hunt for the bad card lasted
two days :( -removed that one all was fine.
But the current failure won't bring down a whole hub, just more failures
will follow, usually for the mostly-engaged cards, so it doesn't seem a
matter of a single card doing the mess, rather a stress situation.
Thanks
_______________________________________________
Pcsclite-muscle mailing list
[email protected]
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pcsclite-muscle
