Daniel Montero Motilla wrote:
Hi, I'm experimenting a strange behaviour using back-meta + slapo-pcache + glue, I have a configuration similar to this:
database meta subordinate advertise suffix "dc=ext,dc=blabla,dc=com" rootdn "cn=Manager,dc=blabla,dc=com" overlay pcache proxycache bdb 10000 3 100 100 ... uri "ldap://10.0.0.1/dc=dir1,dc=ext,dc=blabla,dc=com" suffixmassage "dc=dir1,dc=ext,dc=blabla,dc=com" "dc=w2k3,dc=blabla,dc=local" pseudorootdn "cn=Administrator,cn=Users,dc=w2k3,dc=blabla,DC=local" pseudorootpw blabla
database bdb suffix "dc=blabla,dc=com" rootdn "cn=Manager,dc=blabla,dc=com" rootpw blabla
If I do a search under "dc=dir1,dc=blabla,dc=com" and this search is not in cache, everything works ok, but if the search is obtained from cache, I obtain two different answers to the same 'search' request from slapd, one of them with result code '0' (success) and another one with code 53 (unwilling to perform), although I also obtain the correct result entries. Digging a bit with gdb, I see that the 'unwilling to perform' error is sended in backglue.c (glue_op_search function). Am I maybe doing something wrong on the configuration or is that a normal behaviour?
Looks like the pcache overlay and backglue are both sending results back to the client. This is probably a bug in the pcache overlay.