*This message has been posted to users list but get no responses. Hope here I can get some feedback and solve our critical problem.*
Hi, all! We at Yandex create the new project and use CouchDB as our primary storage. We live in the python universe so use python as client-code language and for view-server. There is number of ready python libraries for CouchDB and we have chosen -- couchdb-python (http://code.google.com/p/couchdb-python/). CouchDB is hosted on Ubuntu 10.04 boxes (64bit; 16 Cores; 23GB RAM) with master + 2 slaves setup. And databases from number of megabytes to some gigabytes in size. We faced some strange CouchDB behavior which is related to our view-server and its integration with storage (I think). We get `timeout` errors from the storage. This errors appear even when trying to issue a replication, getting database information (/db/_info) or querying views itself. Here is some logs with error messages -- http://dpaste.com/hold/258268/ . View-server works fine and has no error messages. But, we've noticed that it is getting some strange `reduce` requests with empty arguments. Views are very simple and cannot process input so long. Our local.ini has `os_process_tmeout` preset with 6000000 (six million milliseconds) and it doesn't help. It seams that this timeout doesn't event count and response with `timeout` error returns in some seconds (over 5 I think). The same issue we've encountered earlier with Ubuntu 8.04 (Hardy) and previous version (pre 1.0) of CouchDB. What do you think about that? Where is our mistake or misconfiguration? Thanks! --- Alexander Koshelev
