imptcp: Plain TCP Syslog¶
|Author:||Rainer Gerhards <firstname.lastname@example.org>|
Provides the ability to receive syslog messages via plain TCP syslog. This is a specialised input plugin tailored for high performance on Linux. It will probably not run on any other platform. Also, it does not provide TLS services. Encryption can be provided by using stunnel.
This module has no limit on the number of listeners and sessions that can be used.
Parameter names are case-insensitive.
Number of helper worker threads to process incoming messages. These threads are utilized to pull data off the network. On a busy system, additional helper threads (but not more than there are CPUs/Cores) can help improving performance. The default value is two, which means there is a default thread count of three (the main input thread plus two helpers). No more than 16 threads can be set (if tried to, rsyslog always resorts to 16).
Instructs imptcp to process messages on poller thread opportunistically. This leads to lower resource footprint(as poller thread doubles up as message-processing thread too). “On” works best when imptcp is handling low ingestion rates.
At high throughput though, it causes polling delay(as poller spends time processing messages, which keeps connections in read-ready state longer than they need to be, filling socket-buffer, hence eventually applying backpressure).
It defaults to allowing messages to be processed on poller (for backward compatibility).
These parameters can be used with the “input()” statement. They apply to the input they are specified with.
Select a port to listen on. It is an error to specify both path and port.
A path on the filesystem for a unix domain socket. It is an error to specify both path and port.
When a message is split because it is to long the second part is normally processed as the next message. This can cause Problems. When this parameter is turned on the part of the message after the truncation will be discarded.
Set the file owner for the domain socket. The parameter is a user name, for which the userid is obtained by rsyslogd during startup processing. Interim changes to the user mapping are not detected.
Set the file owner for the domain socket. The parameter is a numerical ID, which which is used regardless of whether the user actually exists. This can be useful if the user mapping is not available to rsyslog during startup.
Set the group for the domain socket. The parameter is a group name, for which the groupid is obtained by rsyslogd during startup processing. Interim changes to the user mapping are not detected.
Set the group for the domain socket. The parameter is a numerical ID, which is used regardless of whether the group actually exists. This can be useful if the group mapping is not available to rsyslog during startup.
Set the access permissions for the domain socket. The value given must always be a 4-digit octal number, with the initial digit being zero. Please note that the actual permission depend on rsyslogd’s process umask. If in doubt, use “$umask 0000” right at the beginning of the configuration file to remove any restrictions.
Rsyslog will not start if this is on and changing the file owner, group, or access permissions fails. Disable this to ignore these errors.
Sets a name for the inputname property. If no name is set “imptcp” is used by default. Setting a name is not strictly necessary, but can be useful to apply filtering based on which input the message was received from. Note that the name also shows up in impstats logs.
Binds specified ruleset to this input. If not set, the default ruleset is bound.
When in octet counted mode, the frame size is given at the beginning of the message. With this parameter the max size this frame can have is specified and when the frame gets to large the mode is switched to octet stuffing. The max value this parameter can have was specified because otherwise the integer could become negative and this would result in a Segmentation Fault. (Max Value: 200000000)
On multi-homed machines, specifies to which local address the listener should be bound.
This directive permits to specify an additional frame delimiter for plain tcp syslog. The industry-standard specifies using the LF character as frame delimiter. Some vendors, notable Juniper in their NetScreen products, use an invalid frame delimiter, in Juniper’s case the NUL character. This directive permits to specify the ASCII value of the delimiter in question. Please note that this does not guarantee that all wrong implementations can be cured with this directive. It is not even a sure fix with all versions of NetScreen, as I suggest the NUL character is the effect of a (common) coding error and thus will probably go away at some time in the future. But for the time being, the value 0 can probably be used to make rsyslog handle NetScreen’s invalid syslog/tcp framing. For additional information, see this forum thread. If this doesn’t work for you, please do not blame the rsyslog team. Instead file a bug report with Juniper!
Note that a similar, but worse, issue exists with Cisco’s IOS implementation. They do not use any framing at all. This is confirmed from Cisco’s side, but there seems to be very limited interest in fixing this issue. This directive can not fix the Cisco bug. That would require much more code changes, which I was unable to do so far. Full details can be found at the Cisco tcp syslog anomaly page.
The legacy octed-counted framing (similar to RFC5425 framing) is activated. This is the default and should be left unchanged until you know very well what you do. It may be useful to turn it off, if you know this framing is not used and some senders emit multi-line messages into the message stream.
Instructs imptcp to emit a message if a remote peer closes the connection.
Instructs imptcp to emit a message if a remote peer opens a connection. Hostname of the remote peer is given in the message.
Enable of disable keep-alive packets at the tcp socket layer. The default is to disable them.
The number of unacknowledged probes to send before considering the connection dead and notifying the application layer. The default, 0, means that the operating system defaults are used. This has only effect if keep-alive is enabled. The functionality may not be available on all platforms.
The interval between subsequential keepalive probes, regardless of what the connection has exchanged in the meantime. The default, 0, means that the operating system defaults are used. This has only effect if keep-alive is enabled. The functionality may not be available on all platforms.
The interval between the last data packet sent (simple ACKs are not considered data) and the first keepalive probe; after the connection is marked to need keepalive, this counter is not used any further. The default, 0, means that the operating system defaults are used. This has only effect if keep-alive is enabled. The functionality may not be available on all platforms.
Specifies the rate-limiting interval in seconds. Set it to a number of seconds (5 recommended) to activate rate-limiting.
Specifies the rate-limiting burst in number of messages.
This is the counterpart to the compression modes set in omfwd. Please see it’s documentation for details.
Flow control is used to throttle the sender if the receiver queue is near-full preserving some space for input that can not be throttled.
Experimental parameter which causes rsyslog to recognise a new message only if the line feed is followed by a ‘<’ or if there are no more characters.
Experimental parameter. It is similar to “MultiLine”, but provides greater control of when a log message ends. You can specify a regular expression that characterizes the header to expect at the start of the next message. As such, it indicates the end of the current message. For example, one can use this setting to use a RFC3164 header as frame delimiter:
Note that when oversize messages arrive this mode may have problems finding the proper frame terminator. There are some provisions inside imptcp to make these kinds of problems unlikely, but if the messages are very much over the configured MaxMessageSize, imptcp emits an error messages. Chances are great it will properly recover from such a situation.
Specifies the backlog parameter sent to the listen() function. It defines the maximum length to which the queue of pending connections may grow. See man page of listen(2) for more information. The parameter controls both TCP and UNIX sockets backlog parameter. Default value is arbitrary set to 5.
Set default time zone. At most seven chars are set, as we would otherwise overrun our buffer.
Cisco very occasionally sends a space after a line feed, which thrashes framing if not taken special care of. When this parameter is set to “on”, we permit space in front of the next frame and ignore it.
New in version 8.38.0.
With this parameter you can specify the name for a file. In this file the port, imptcp is connected to, will be written. This parameter was introduced because the testbench works with dynamic ports.
This plugin maintains statistics for each listener. The statistic is named “imtcp” , followed by the bound address, listener port and IP version in parenthesis. For example, the counter for a listener on port 514, bound to all interfaces and listening on IPv6 is called “imptcp(*/514/IPv6)”.
The following properties are maintained for each listener:
- submitted - total number of messages submitted for processing since startup
When a message is to long it will be truncated and an error will show the remaining length of the message and the beginning of it. It will be easier to comprehend the truncation.
- module always binds to all interfaces
This sets up a TCP server on port 514:
module(load="imptcp") # needs to be done just once input(type="imptcp" port="514")
This creates a listener that listens on the local loopback interface, only.
module(load="imptcp") # needs to be done just once input(type="imptcp" port="514" address="127.0.0.1")
Create a unix domain socket:
module(load="imptcp") # needs to be done just once input(type="imptcp" path="/tmp/unix.sock" unlink="on")
Help with configuring/using