ctdb-statistics
7
ctdb
CTDB - clustered TDB database
ctdb-statistics
CTDB statistics output
OVERALL STATISTICS
CTDB maintains information about various messages communicated
and some of the important operations per node. See the
ctdb
1 commands
statistics and statisticsreset
for displaying statistics.
Example: ctdb statistics
CTDB version 1
Current time of statistics : Fri Sep 12 13:32:32 2014
Statistics collected since : (000 01:49:20) Fri Sep 12 11:43:12 2014
num_clients 6
frozen 0
recovering 0
num_recoveries 2
client_packets_sent 281293
client_packets_recv 296317
node_packets_sent 452387
node_packets_recv 182394
keepalive_packets_sent 3927
keepalive_packets_recv 3928
node
req_call 48605
reply_call 1
req_dmaster 23404
reply_dmaster 24917
reply_error 0
req_message 958
req_control 197513
reply_control 153705
client
req_call 130866
req_message 770
req_control 168921
timeouts
call 0
control 0
traverse 0
locks
num_calls 220
num_current 0
num_pending 0
num_failed 0
total_calls 130866
pending_calls 0
childwrite_calls 1
pending_childwrite_calls 0
memory_used 334490
max_hop_count 18
total_ro_delegations 2
total_ro_revokes 2
hop_count_buckets: 42816 5464 26 1 0 0 0 0 0 0 0 0 0 0 0 0
lock_buckets: 9 165 14 15 7 2 2 0 0 0 0 0 0 0 0 0
locks_latency MIN/AVG/MAX 0.000685/0.160302/6.369342 sec out of 214
reclock_ctdbd MIN/AVG/MAX 0.004940/0.004969/0.004998 sec out of 2
reclock_recd MIN/AVG/MAX 0.000000/0.000000/0.000000 sec out of 0
call_latency MIN/AVG/MAX 0.000006/0.000719/4.562991 sec out of 126626
childwrite_latency MIN/AVG/MAX 0.014527/0.014527/0.014527 sec out of 1
CTDB version
Version of the ctdb protocol used by the node.
Current time of statistics
Time when the statistics are generated.
This is useful when collecting statistics output periodically
for post-processing.
Statistics collected since
Time when ctdb was started or the last time statistics was reset.
The output shows the duration and the timestamp.
num_clients
Number of processes currently connected to CTDB's unix socket.
This includes recovery daemon, ctdb tool and samba processes
(smbd, winbindd).
frozen
1 if the the databases are currently frozen, 0 otherwise.
recovering
1 if recovery is active, 0 otherwise.
num_recoveries
Number of recoveries since the start of ctdb or since the last
statistics reset.
client_packets_sent
Number of packets sent to client processes via unix domain socket.
client_packets_recv
Number of packets received from client processes via unix domain socket.
node_packets_sent
Number of packets sent to the other nodes in the cluster via TCP.
node_packets_recv
Number of packets received from the other nodes in the cluster via TCP.
keepalive_packets_sent
Number of keepalive messages sent to other nodes.
CTDB periodically sends keepalive messages to other nodes.
See KeepaliveInterval tunable in
ctdb-tunables
7 for more details.
keepalive_packets_recv
Number of keepalive messages received from other nodes.
node
This section lists various types of messages processed which
originated from other nodes via TCP.
req_call
Number of REQ_CALL messages from the other nodes.
reply_call
Number of REPLY_CALL messages from the other nodes.
req_dmaster
Number of REQ_DMASTER messages from the other nodes.
reply_dmaster
Number of REPLY_DMASTER messages from the other nodes.
reply_error
Number of REPLY_ERROR messages from the other nodes.
req_message
Number of REQ_MESSAGE messages from the other nodes.
req_control
Number of REQ_CONTROL messages from the other nodes.
reply_control
Number of REPLY_CONTROL messages from the other nodes.
client
This section lists various types of messages processed which
originated from clients via unix domain socket.
req_call
Number of REQ_CALL messages from the clients.
req_message
Number of REQ_MESSAGE messages from the clients.
req_control
Number of REQ_CONTROL messages from the clients.
timeouts
This section lists timeouts occurred when sending various messages.
call
Number of timeouts for REQ_CALL messages.
control
Number of timeouts for REQ_CONTROL messages.
traverse
Number of timeouts for database traverse operations.
locks
This section lists locking statistics.
num_calls
Number of completed lock calls. This includes database locks
and record locks.
num_current
Number of scheduled lock calls. This includes database locks
and record locks.
num_pending
Number of queued lock calls. This includes database locks and
record locks.
num_failed
Number of failed lock calls. This includes database locks and
record locks.
total_calls
Number of req_call messages processed from clients. This number
should be same as client --> req_call.
pending_calls
Number of req_call messages which are currenly being processed.
This number indicates the number of record migrations in flight.
childwrite_calls
Number of record update calls. Record update calls are used to
update a record under a transaction.
pending_childwrite_calls
Number of record update calls currently active.
memory_used
The amount of memory in bytes currently used by CTDB using
talloc. This includes all the memory used for CTDB's internal
data structures. This does not include the memory mapped TDB
databases.
max_hop_count
The maximum number of hops required for a record migration request
to obtain the record. High numbers indicate record contention.
total_ro_delegations
Number of readonly delegations created.
total_ro_revokes
Number of readonly delegations that were revoked. The difference
between total_ro_revokes and total_ro_delegations gives the
number of currently active readonly delegations.
hop_count_buckets
Distribution of migration requests based on hop counts values.
Buckets are 1, < 4, < 8, < 16, < 32, < 64, <
128, < 256, < 512, ≥ 512.
lock_buckets
Distribution of record lock requests based on time required to
obtain locks. Buckets are < 1ms, < 10ms, < 100ms,
< 1s, < 2s, < 4s, < 8s, < 16s, < 32s, <
64s, ≥ 64s.
locks_latency
The minimum, the average and the maximum time (in seconds)
required to obtain record locks.
reclock_ctdbd
The minimum, the average and the maximum time (in seconds)
required to check if recovery lock is still held by recovery
daemon when recovery mode is changed. This check is done in ctdb daemon.
reclock_recd
The minimum, the average and the maximum time (in seconds)
required to check if recovery lock is still held by recovery
daemon during recovery. This check is done in recovery daemon.
call_latency
The minimum, the average and the maximum time (in seconds) required
to process a REQ_CALL message from client. This includes the time
required to migrate a record from remote node, if the record is
not available on the local node.
childwrite_latency
Default: 0
The minimum, the average and the maximum time (in seconds)
required to update records under a transaction.
DATABASE STATISTICS
CTDB maintains per database statistics about important operations.
See the ctdb
1 command
dbstatistics for displaying database statistics.
Example: ctdb dbstatistics notify_index.tdb
DB Statistics: notify_index.tdb
ro_delegations 0
ro_revokes 0
locks
total 131
failed 0
current 0
pending 0
hop_count_buckets: 9890 5454 26 1 0 0 0 0 0 0 0 0 0 0 0 0
lock_buckets: 4 117 10 0 0 0 0 0 0 0 0 0 0 0 0 0
locks_latency MIN/AVG/MAX 0.000683/0.004198/0.014730 sec out of 131
Num Hot Keys: 3
Count:7 Key:2f636c75737465726673
Count:18 Key:2f636c757374657266732f64617461
Count:7 Key:2f636c757374657266732f646174612f636c69656e7473
DB Statistics
Name of the database.
ro_delegations
Number of readonly delegations created in the database.
ro_revokes
Number of readonly delegations revoked. The difference in
ro_delegations and ro_revokes indicates the currently active
readonly delegations.
locks
This section lists locking statistics.
total
Number of completed lock calls. This includes database locks
and record locks.
failed
Number of failed lock calls. This includes database locks and
record locks.
current
Number of scheduled lock calls. This includes database locks
and record locks.
pending
Number of queued lock calls. This includes database locks and
record locks.
hop_count_buckets
Distribution of migration requests based on hop counts values.
Buckets are 1, < 4, < 8, < 16, < 32, < 64, <
128, < 256, < 512, ≥ 512.
lock_buckets
Distribution of record lock requests based on time required to
obtain locks. Buckets are < 1ms, < 10ms, < 100ms,
< 1s, < 2s, < 4s, < 8s, < 16s, < 32s, <
64s, ≥ 64s.
locks_latency
The minimum, the average and the maximum time (in seconds)
required to obtain record locks.
Num Hot Keys
Number of contended records determined by hop count. CTDB keeps
track of top 10 hot records and the output shows hex encoded
keys for the hot records.
SEE ALSO
ctdb
1,
ctdbd
1,
ctdb-tunables
7,
This documentation was written by
Amitay Isaacs,
Martin Schwenke
2007
Andrew Tridgell
Ronnie Sahlberg
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 3 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be
useful, but WITHOUT ANY WARRANTY; without even the implied
warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public
License along with this program; if not, see
.