The minOpTime recovery record is currently written regardless of the type of the catalog manager. This causes problems at a later time, when the shard is being restored, because it will try to instantiate a legacy catalog manager, even though it might not be in use.
Example log output:
2015-10-30T17:03:46.552+0000 I CONTROL [initandlisten] MongoDB starting : pid=31310 port=27503 dbpath=/data/backups/daemon/544e5cb9e4b00ae3893aaa70/cluster_test_1/head/ 64-bit host=mms-qa-daemon-1 2015-10-30T17:03:46.553+0000 I CONTROL [initandlisten] db version v3.2.0-rc1 2015-10-30T17:03:46.553+0000 I CONTROL [initandlisten] git version: beabb900fa05c3b090fc62e887d41d9c43562b3f 2015-10-30T17:03:46.553+0000 I CONTROL [initandlisten] OpenSSL version: OpenSSL 1.0.1c 10 May 2012 2015-10-30T17:03:46.553+0000 I CONTROL [initandlisten] allocator: tcmalloc 2015-10-30T17:03:46.553+0000 I CONTROL [initandlisten] modules: none 2015-10-30T17:03:46.553+0000 I CONTROL [initandlisten] build environment: 2015-10-30T17:03:46.553+0000 I CONTROL [initandlisten] distmod: ubuntu1204 2015-10-30T17:03:46.553+0000 I CONTROL [initandlisten] distarch: x86_64 2015-10-30T17:03:46.553+0000 I CONTROL [initandlisten] target_arch: x86_64 2015-10-30T17:03:46.553+0000 I CONTROL [initandlisten] options: { net: { bindIp: "127.0.0.1", port: 27503 }, operationProfiling: { slowOpThresholdMs: 1000 }, setParameter: { failIndexKeyTooLong: "false", ttlMonitorEnabled: "false" }, storage: { dbPath: "/data/backups/daemon/544e5cb9e4b00ae3893aaa70/cluster_test_1/head/", engine: "mmapv1", journal: { enabled: false } }, systemLog: { destination: "file", logAppend: true, path: "/data/backups/daemon/544e5cb9e4b00ae3893aaa70/cluster_test_1/mongod.log", quiet: true } } 2015-10-30T17:03:46.586+0000 I FTDC [initandlisten] Starting full-time diagnostic data capture with directory '/data/backups/daemon/544e5cb9e4b00ae3893aaa70/cluster_test_1/head/diagnostic.data' 2015-10-30T17:03:46.586+0000 I NETWORK [HostnameCanonicalizationWorker] Starting hostname canonicalization worker 2015-10-30T17:03:46.586+0000 I SHARDING [initandlisten] Sharding state recovery process found document { _id: "minOpTimeRecovery", configsvrConnectionString: "keylime:30005", shardName: "cluster_test_1", minOpTime: { ts: Timestamp 0|0, t: -1 }, minOpTimeUpdaters: 0 } 2015-10-30T17:03:46.586+0000 I SHARDING [initandlisten] first cluster operation detected, adding sharding hook to enable versioning and authentication to remote servers 2015-10-30T17:03:46.586+0000 I SHARDING [initandlisten] Updating config server connection string to: keylime:30005 2015-10-30T17:03:46.589+0000 I NETWORK [initandlisten] getaddrinfo("keylime") failed: Name or service not known 2015-10-30T17:03:46.589+0000 I SHARDING [initandlisten] can't resolve DNS for [keylime] sleeping and trying 10 more times 2015-10-30T17:03:56.590+0000 I NETWORK [initandlisten] getaddrinfo("keylime") failed: Name or service not known 2015-10-30T17:03:56.590+0000 I SHARDING [initandlisten] can't resolve DNS for [keylime] sleeping and trying 9 more times 2015-10-30T17:04:06.591+0000 I NETWORK [initandlisten] getaddrinfo("keylime") failed: Name or service not known 2015-10-30T17:04:06.591+0000 I SHARDING [initandlisten] can't resolve DNS for [keylime] sleeping and trying 8 more times 2015-10-30T17:04:16.592+0000 I NETWORK [initandlisten] getaddrinfo("keylime") failed: Name or service not known 2015-10-30T17:04:16.592+0000 I SHARDING [initandlisten] can't resolve DNS for [keylime] sleeping and trying 7 more times 2015-10-30T17:04:26.592+0000 I NETWORK [initandlisten] getaddrinfo("keylime") failed: Name or service not known 2015-10-30T17:04:26.593+0000 I SHARDING [initandlisten] can't resolve DNS for [keylime] sleeping and trying 6 more times 2015-10-30T17:04:36.593+0000 I NETWORK [initandlisten] getaddrinfo("keylime") failed: Name or service not known 2015-10-30T17:04:36.593+0000 I SHARDING [initandlisten] can't resolve DNS for [keylime] sleeping and trying 5 more times 2015-10-30T17:04:42.371+0000 I CONTROL [signalProcessingThread] got signal 15 (Terminated), will terminate after current cmd ends 2015-10-30T17:04:42.371+0000 I FTDC [signalProcessingThread] Stopping full-time diagnostic data capture 2015-10-30T17:04:46.595+0000 I NETWORK [initandlisten] getaddrinfo("keylime") failed: Name or service not known 2015-10-30T17:04:46.596+0000 I SHARDING [initandlisten] can't resolve DNS for [keylime] sleeping and trying 4 more times 2015-10-30T17:04:56.596+0000 I NETWORK [initandlisten] getaddrinfo("keylime") failed: Name or service not known 2015-10-30T17:04:56.596+0000 I SHARDING [initandlisten] can't resolve DNS for [keylime] sleeping and trying 3 more times 2015-10-30T17:05:06.597+0000 I NETWORK [initandlisten] getaddrinfo("keylime") failed: Name or service not known 2015-10-30T17:05:06.597+0000 I SHARDING [initandlisten] can't resolve DNS for [keylime] sleeping and trying 2 more times 2015-10-30T17:05:16.598+0000 I NETWORK [initandlisten] getaddrinfo("keylime") failed: Name or service not known 2015-10-30T17:05:16.598+0000 I SHARDING [initandlisten] can't resolve DNS for [keylime] sleeping and trying 1 more times 2015-10-30T17:05:26.599+0000 I STORAGE [initandlisten] exception in initAndListen: 7 unable to resolve DNS for host keylime, terminating