After SSO (chocolatechip) cookie validation on the slave the SSO cookie is resent, this time with cookie['master'] equal to FALSE (because it's set on the slave).
It's difficult to define in reproducible steps, but should authentication fail on the subsite, and the SSO cookie is reset, then the next requests will never authenticate, because the SSO cookie will no longer validate due to having been generated on the slave.
| Comment | File | Size | Author |
|---|---|---|---|
| #1 | Bakery-sso-cookie-1278168.patch | 998 bytes | coltrane |
| #1 | Bakery-sso-cookie-d6-1278168.patch | 998 bytes | coltrane |
Comments
Comment #1
coltraneHere's a possible patch for D6 and D7. I've tested a little, but I'm also trying to narrow down the steps for the bug. Sometimes it seems infrastructure specific (Varnish caching etc.) so difficult to generalize and reproduce in my local vhost setup.
Comment #2
purencool commented