Note that lots of the info that should be on this page has not been written up yet. Please see the to do list for information on how you can help.
At startup, dtnd creates an initializes a TCL interpreter. It then loads and evaluates the contents of the configuration file. For more info on the TCL language, see the TCL homepage.
The name of the configuration file is usually set with the -c argument. If the -c argument is missing, dtnd will try to load /etc/dtn.conf. If that file is not readable, it will try to read daemon/dtn.conf. If that file is not readable, it will log a warning and continue without a configuration file. You may still configure dtnd by hand at the dtn% prompt.
Each of the commands that dtnd handles is described below. These are extensions to the normal TCL command set. You may always use TCL commands and other expressions in your dtn.conf configuration file. For an example of this, search for [info hostname] in the sample daemon/dtn.conf file.
Syntax: api set variable value
Example: api set local_addr 10.0.0.1
Use the api command to configure the interface between dtnd and the DTN applications (like dtnping, dtncp, etc). See also environment variables which are used to control the behavior of the API clients.
Variable | Possible settings | Default | Comments |
---|---|---|---|
local_addr | An IP address or a hostname. | 127.0.0.1 | The IP address the API Server will listen to for requests from API clients. |
local_port | An IP port number | 5010 | The IP port on which the API Server listens for requests from API clients. |
You can use the bundle command to see the status of bundles currently being held by your dtnd.
Syntax: bundle command [id]
Example: storage set type berkeleydb
Use the bundle command to get high level information on the bundle and bundle daemon statistcs,
Command | Comments |
---|---|
stats | list bundle statistics. |
daemon_stats | list bundle daemon statistics. |
reset_stats | reset bundle statistics. |
list | list all bundle. |
ids | list all bundle IDs. |
While testing a DTN, you can also inject bundles.
Syntax: bundle inject src dest payload opt1=val1, opt2=val2 optN=valN...
Example: bundle inject dtn://source.dtn dtn://dest.dtn payload
The bundle stats command shows the following counts:
Pending is the current count of bundles in dtnd's store, which may be non-zero when the server starts if bundles are still in the store from the last time the server was running. All the other counters are zeroed when the server boots. Pending is practically limited by the storage system and the memory use of dtnd. The other counters are unsigned 32-bit integers, and will roll over accordingly.
bundle reset_stats may be used to reset currently maintained statistics.
The bundle daemon_stats command shows the following counts:
The bundle list command shows a list of all pending bundles in the system. It looks like this:
dtn% bundle list Currently Pending Bundles (2): 0 : dtn://mars/rover -> dtn://jpl/images length 8096266 1 : dtn://jpl/commands -> dtn://mars/rover length 18
To the left of the colon is the bundle ID. This number is used internally by dtnd to keep track of bundles. Remote DTN implementations will never see the bundle ID. However, you can use the bundle ID with the following commands to inspect the bundles.
Using bundle info bundle-id, you can see more information about a bundle. Using bundle dump bundle-id, you can see a hex dump of the bundle. If you know the bundle has only printable characters in it, you can see it formatted as a text file using bundle dump_ascii bundle-id.
Syntax: bundle inject source dest payload
[ length ]
Example: bundle inject dtn://jpl/commands
dtn://mars/rover "turn left 30 degrees, please"
You can use bundle inject to inject a synthetic bundle into dtnd. It will route and deliver that packet just like any other bundle, as long as it is valid. You have more control over bundles you add using the dtnsend client application, so it is probably almost always a better choice, even for testing situations.
Note: It is currently possible to crash the server using bundle inject with incorrect parameters. Please do not use this command unless you can tolerate a server crash. The crash comes from a failing assert, so the daemon notices the bad bundle, it just cannot proceed with the bad bundle in the store. Once it the bad bundle is in the store, you will need to use the "--tidy" option to remove it and any other pending bundles.
The length parameter is optional. If you inject a bundle without using it, the length of the bundle will simply be the length of the string you pass in as the payload. The length parameter can allow you to make huge bundles without having to use a huge payload string. The payload string is used for at the beginning of the bundle and the remainder of the bundle is filled with zeros. (But there's a bug that makes this not work right. Try dumping out a bundle you made this way to see an assert.)
The console command is used spawn interpreter on stdin/stdout, set the console prompt string and set the console listening address and port number.
Syntax: console set arg
Example: console set prompt "$shorthostname dtn% "
Variable | Possible settings | Default | Comments |
---|---|---|---|
stdio | None | stdio | spawn interpreter on stdin/stdout |
addr | A.B.C.D | 127.0.0.1 | console listening address |
port | number | 0 | console listening port |
prompt | string | shorthostname [lindex [split [info hostname] .] 0] dtn% | console prompt string |
The debug command exists in all versions of dtnd, but the subcommands are only available if you gave "--enable-debug-memory" to the configure script while building dtnd. It is not enabled by default. You can tell if your dtnd binary has memory debugging support by running help debug.
When memory use debugging is enabled, you can get a dump of memory in use using debug dump_memory. After doing one dump, you can get diffs with debug dump_memory_diffs.
For information on debugging and configuring logging of the dtn daemon please see the logging section and .dtndebug
The discovery command is used to configure the abstraction of a neighbor discovery agent. Much like Interface, Discovery is generally created by the configuration file / console. Derived classes (such as IPDiscovery) typically bind to a UDP socket to listen for neighbor beacons. Bluetooth has built-in discovery mechanisms, so BluetoothDiscovery polls via Inquiry instead of listening on a socket.
To advertise a local convergence layer, register its local address (and port) by calling "discovery add". For each registered CL, Discovery will advertise (outbound) the CL's presence to neighbors,and distribute (inbound) each event of neighbor discovery to each CL.
This command is used to add a discovery agent.
Syntax: discovery add discovery_name cl-type [ args ]
Example: discovery add udp0d ip port=9556
Example : discovery add btd bt
Example: discovery add bonjourd bonjour
Variable | Possible settings | Comments |
---|---|---|
discovery_name | string | A name string to identify the discovery agent |
cl-type | bt | ip | bonjour | The convergence layer type. |
args | Possible settings | Comments |
---|---|---|
port | number | Discovery port number if IP based |
continue_on_error | true or false | Whether to continue on an error |
addr | A.B.C.D | IPv4 address of .. |
local_addr | A.B.C.D | IPv4 address of .. |
multicast_ttl | number | Multicast time to live |
unicast | true or false | Whether is unicast |
This command is used to announce the address of a local interface (convergence layer).
Syntax: discovery announce cl-name discovery_name cl-type
Example discovery announce udp0 udp0d udp interval=10
Example discovery announce bt0 btd bt interval=1
args | Possible settings | Comments |
---|---|---|
cl_addr | A.B.C.D | IPv4 address of CL to announce |
cl_port | number (port) | port number of CL to announce |
This command is used to remove an announcement for a local interface.
Syntax: discovery remove announce discovery_name
Example: discovery remove bt0 btd
This command is used to delete a discovery agent.
Syntax: discovery del discovery_name
This command is used to list agents and their announcement parameters.
Syntax: discovery list
The ecla command is used to control the external convergence layer (ECL) is the gateway between DTN2 and convergence layer adapters (CLAs) running outside of DTN2. Please see ecla for details.
Syntax: ecla command
Example: ecla set create_discovered_links
The gettimeofday command is used to print the result of gettimeofday() in secs.usecs format.
Syntax: gettimeofday
Example: gettimeofday
You can get help on any of the additional TCL commands in dtnd's TCL interpreter using the help command.
Syntax: help command
Example: help interface
Command | Comments |
---|---|
bundle | print bundle and bundle daemon statistics and interact with the bundle daemon. Also used to examine, dump, expire, cancel and inject bundles. |
console | set console parameters |
debug | Command for memory debugging. Only available if "--enable-debug-memory" was given to the configure script. |
discovery | add, delete and list discovery agent and announce parameters |
ecla | list and set the external convergence layer adapter parameters |
gettimeofday | print the result of gettimeofday() in secs.usecs format. |
help | print the help documentation for cmd. |
interface | add, delete and list interface parameters. |
link | add, delete, reconfigure and list link parameters and statistics. Also used to modify default CLA parameters. |
log | set log and debug file pathname and level and prefix. Also used to rotate log file, reparse rule file and list filter rules. |
param | set various dtnd daemon parameters and policy. |
prophet | manipulate the prophet routing protocol. These parameters should be manipulated under the route command interface like other DTN routing algorithms and protocols. |
quit | shutdown the daemon |
registration | add, delete and list endpoint (EID) registrations and modify registration logger. |
route | add, delete, list routes, select and configure routing algorithms/protocols. Can be used to set various route parameters including setting of type, next hop routes, route priority, interaction with discovery, DTLSR parameters, and external route parameters. |
shutdown | gracefully shutdown the dtnd daemon. |
storage | set storage backend type, parameters and view statistics. |
test | Command for developer tests |
Use the interface command to add and delete interfaces to the networks that your dtnd will talk on.
Syntax: interface add interface-name convergence-layer [ args ]
Example: interface add tcp0 tcp
Example: interface add udp0 udp local_host=1.2.3.4 local_port=9999
Example: interface add bt0 bt channel=10
Example: interface add eth string://eth0 beacon_interval=1
Example: interface add null
When dtnd encounters an interface add command, it checks to make sure the named interface has not been created before. Next, it looks for a matching convergence layer. If both of those steps succeed, it passes the remaining arguments to the convergence layer to configure a new interface which will listen for incoming bundles.
The arguments, if they exist, are a list of strings (delimited by "",
or {}, according to TCL syntax). Each string starts with the parameter
name, then an equal sign ("="), then the value.
See the following pages on the different convergence layers,
and what arguments they take.
Syntax: interface del convergence-layer
Example: interface del tcp0
Deletes an interface from the system. Each convergence layer handles this individually, but generally the convergence layer stops listening for new bundles and cleans up any resources it was using.
Syntax: interface list
Example: interface list
Prints a list of the current interfaces.
The log command deals with the log file, and the log messages in it. You can add messages of your own to the log file, or trigger various actions related to the log file.
Syntax: log log-path log-level log-string
Example: log /dtnd info "dtnd parsing configuration..."
You can use the log command to add a log message to the dtnd log file. The log-path and log-level are used to filter out messages, as described in the section on Logging.
Syntax: log prefix prefix
Example: log prefix testing
The log prefix shows up in log entries after the left bracket, and
before the timestamp. You can set it to any string.
Syntax: log rotate
Use this command to force dtnd to reopen the log file for writing.
Under Unix, even after you have renamed the log file, dtnd will
continue writing log entries into the file until it executes the
log rotate command.
You can include a procedure like this in your dtnd.conf file
to arrange for dtnd to make a new log file every hour:
Of course, for this to work, you need to start dtnd in the directory
where the log file will be, and use the "-o dtnd.log" argument to set
the log file name.
Syntax: log reparse_debug_file
The system which dtnd uses to choose which messages to log and
which to hide is described in the Logging
section of the manual. This command forces dtnd to re-read the
.dtndebug file and take into account any new rules that it
finds for showing or hiding messages.
The param command is used to configure policy parameters.
Syntax: param set arg val
Rotate the log file
Example: log rotate
# Rotate the log every 3600 seconds (1 hour)
# the initial call to rotate (after the proc) is required to
# start the proc running every hour.
proc rotate {} {
set date [ clock format [ clock seconds ] -format "%Y-%m-%d-%H" ]
set new "dtnd.$date.log"
file rename "dtnd.log" $new
log rotate
after 3600000 rotate
}
rotate
Change log level
Example: log reparse_debug_file
param
param set
Example: param set link_min_retry_interval 30
arg | Possible settings | Default | Comments |
---|---|---|---|
payload_test_no_remove | true or false | false | Whether not to remove bundles (for testing). |
early_deletion | true or false | true | Whether to delete forwarded / delivered bundles before they've expired |
suppress_duplicates | true or false | true | Whether to not route bundles that are a duplicate of any currently pending bundle |
accept_custody | true or false | true | Whether to accept custody when requested |
reactive_frag_enabled | true or false | true | Whether reactive fragmentation is enabled |
retry_reliable_unacked | true or false | true | Whether to retry unacked transmissions on reliable CLs |
test_permuted_delivery | true or false | false | Whether to permute the order of bundles before delivering to registrations. |
injected_bundles_in_memory | true or false | false | Whether injected bundles are held in memory by default. |
is_singleton_default | unknown|singleton|multinode | TBC | How to set the is_singleton bit for unknown schemes. |
glob_unknown_schemes | true or false | false | Whether unknown schemes use glob-based matching for registrations and routes. |
link_min_retry_interval | number (interval in seconds) | TBC | Default minimum connection retry interval for links. |
link_max_retry_interval | number (interval in seconds) | TBC | Default maximum connection retry interval for links. |
custody_timer_min | number (seconds) | TBC | Default value for custody timer min. |
custody_timer_lifetime_pct | number (per cent) | TBC | Default value for custody timer lifetime percentage |
custody_timer_max | number (seconds) | TBC | Default value for custody timer max. |
The prophet command is used to configure the prophet router. This command should not be here but under route like all other DTN routing protocol and algorithm commands. Please see prophet router for detailed infomation.
The quit command is used to shutdown the DTN2 daemon. It is the same as shutdown.
A registration is used to represent an "application" endpoint ID registration, loosely defined to also include internal router mechanisms that consume bundles. Registration state is stored persistently in the database.
The registration command is used add, delete and list EID registrations.
Syntax: registration add opt endpoint
Example: registration add tcl dtn:alex
opt | Possible settings | Comments |
---|---|---|
opt | logger or tcl | Whether to add a logger or tcl representation |
endpoint | An EID (URI) | The EID which you wish to add a registration for |
The logger option is used mostly for testing registrations. When created, this sets up a new registration within the daemon, and for any bundles that arrive, outputs logs of the bundle header fields as well as the payload data (if ascii). The implementation is structured as a thread that blocks (forever) waiting for bundles to arrive on the registration's bundle list, then logging the bundles and looping again. The logpath is set to /dtn/reg/logging/.
Syntax: registration del reg_id
Example: registration del 14
Syntax: registration list
Example: registration list
dump_tcl is used to dump a tcl representation of the registration using its registration id.
Syntax: registration dump_tcl reg_id
Example: registration dump_tcl 14
The shutdown command is used to shutdown the DTN2 daemon. It is the same as quit.
Syntax: shutdown
Example: shutdown
Syntax: storage set variable value
Example: storage set type berkeleydb
Use the storage command to configure the manner in which DTN2 will store persistent state like registrations. You may set the following variables:
Variable | Possible settings | Default | Comments |
---|---|---|---|
dbdir | directory | /var/dtn/db | The directory where the DB files will be stored. Only used for type BerkeleyDB. |
dbfile | name | DTN.db | The file name of the BerkeleyDB database. |
payloaddir | directory | /var/dtn/bundles | The directory where bundles in transit will be stored. |
payload_quota | number (bytes) | 0 | storage quota for bundle payloads (0 is unlimited). |
payload_fd_cache_size | number | 32 | number of payload file descriptors to keep open in a cache. |
server_port | number (port) | 0 | TCP port for IPC to external data store. |
schema | name | none | File containing the XML schema for the external data store interface. |
type | berkeleydb, mysql, postgres, filesysdb, memorydb, filesysdb | berkeleydb | Which database backend will be used. |
dbname | name | DB | Set the database name (appended with .db as the filename in berkeleydb, used as-is for SQL variants. |
init_db | true, false | false | When true, dtnd initialize the database on startup. |
tidy | true, false | false | When true, dtnd cleans out the database and the bundle directories at startup. |
tidy_wait | number (seconds) | 3 | How long to wait before really doing a tidy. |
leave_clean_file | true, false | true | Leave a .ds_clean file on clean. |
fs_fd_cache_size | number | 0 | Number of open fds to cache - if > 0 then this # of open fds will be cached when using Filesystem DB. |
db_mpool | true, false | true | Use mpool in Berkeley DB. |
db_log | true, false | true | Use logging in Berkeley DB. |
db_txn | true, false | true | Use transactions in Berkeley DB. |
db_sharefile | true, false | false | Use shared database file in Berkeley DB. |
db_max_tx | number | 0 | Max # of active transactions in Berkeley DB. |
db_max_locks | number | 0 | Max # of active locks in Berkeley DB. |
db_max_lockers | number | 0 | Max # of active locking threads in Berkeley DB. |
db_max_lockedobjs | number | 0 | Max # of active locked objects in Berkeley DB. |
db_lockdetect | number (frequency) | 5000 | Frequency to check for Berkeley DB deadlocks (zero disables locking). |
Tidy is useful when there are pending bundles in the system that you do not want dtnd to continue to try to deliver. If you allow dtnd to tidy up when it has custody of a bundle, that bundle will be lost for all time and the sending application will not be notified.
For SQL backends, it is not possible at this time to configure a hostname, username, or password via the storage command. For MySQL, the defaults are hostname "localhost", Unix user name, and no password. For Postgres, the defaults are taken from the environment variables described in the PostgreSQL manual.
Syntax: route set variable value
Example: route set type static
Use the route command to configure the manner in which DTN2 will route. You may set the following variables:
Variable | Possible settings | Default | Comments |
---|---|---|---|
type | static, prophet, flood, dtlsr, tca_router, tca_gateway, external | static | Which routing algorithm to use. |
add_nexthop_routes | true, false | true | Whether or not to automatically add routes for next hop links . |
add | dest link/endpoint [opts] | N/A | Add a static route |
del | dest link/endpoint | N/A | Delete a static route |
dump | N/A | N/A | print all of the static routes |
open_discovered_links | true, false | true | Whether or not to automatically open discovered opportunistic links. |
default_priority | number | 0 | Default priority for new routes. |
max_route_to_chain | number | 10 | Maximum number of route_to links to follow. |
subscription_timeout | number | 600 | Default timeout for upstream subscription. |
dtlsr_area | string | tbc | DTLSR Administrative area for the local node. |
dtlsr_weight_fn | number | 10 | DTLSR weight function for the graph. |
dtlsr_weight_shift | number | 0 | DTLSR scale factor to scale the weight on down links factor by which to age the cost of a link based on its uptimei. As a link's uptime goes to zero, the cost increases by 10x the original cost. |
dtlsr_uptime_factor | number | 10 | DTLSR aging pct for cost of down links. |
dtlsr_keep_down_links | true or false | true | DTLSR whether or not to keep down links in the graph, marking them as stale. |
dtlsr_recompute_delay | number (seconds) | 1 | DTLSR delay (in seconds) after receiving an LSA when we recompute the routes. Needed to prevent some flapping. |
dtlsr_aging_interval | number (seconds) | 5 | DTLSR interval after which we locally recompute the routes to properly age links that we believe to be down. |
dtlsr_lsa_interval | number (seconds) | 3600 | DTLSR interval at which we proactively send new LSA. Default is once per hour. |
dtlsr_min_lsa_interval | number (seconds) | 5 | DTLSR minimum interval between LSA transmission. Default is once per five seconds. |
dtlsr_lsa_lifetime | number (seconds) | 86400 | DTLSR expiration time for lsa announcements (default is 24 hours). |
server_port | number (port) | 8001 | UDP port for IPC with external router(s) |
hello_interval | number (seconds) | 30 | Seconds between hello messages with external router(s) |
schema | string (pathname) | /router.xsd | Location of the external router interface message schema |
xml_server_validation | true or false | true | Perform xml validation on plug-in interface messages when using external router(s) |
xml_client_validation | true or false | false | Include meta-info in xml messages so plug-in routers can perform validation when using external router(s) |
TCARouter is an implementation of the TCA bundle routing algorithm. A TCARouter is a specialized TableBasedRouter where the route table is manipulated in response to certain control bundles (for example, a "register" bundle, or a "change-of-address" bundle). Specialized routing logic is then applied in order to route late-bound bundles addressed to a mobile node, to the mobile's current location in the network. The main interface point is the overridden handle_bundle_received function which tests for the special TCA bundles (control bundles and late-bound data bundles).
A link is an abstraction for a DTN link, i.e. a one way communication channel to a next hop node in the DTN overlay. Use the link command to configure the manner in which DTN2 will establish a communications channel to a next hop node. You may set the following variables:
Syntax: link cmd name
Example: link delete trinity_od
cmd | name | Comments |
---|---|---|
Open | String | open a link |
close | String | Close a link |
delete | String | Delete a link |
state | String | Get state of a link |
stats | String | Get link statistics |
Syntax: link set_available name avail
Example: link set_available trinity_od true
Variable | Possible settings | Default | Comments |
---|---|---|---|
name | String | N/A | Specify a name of a link |
avail | true or false | true | hacky way to make the link available (only used with set_available) |
Syntax: link reconfigure name opt=val opt2=val2 optN=valN
Example: link reconfigure trinity_od mtu=1024 idle_close_time=35
Syntax: link set_cl_defaults name conv_layer opt=val opt2=val2 optN=valN
Example: link set_cl_defaults trinity_od tcp mtu=1024 idle_close_time=35
Syntax: link add name nexthop type conv_layer opt=val opt2=val2 optN=valN
Example: link add trinity_od 192.168.1.2 ONDEMAND tcp mtu=1024 potential_downtime=25
Variable | Possible settings | Default | Comments |
---|---|---|---|
name | String | N/A | Specify a name of a link |
nexthop | String (address) | N/A | Specify the next hop node. For IP-based links (tcp or udp), the nexthop should contain a DNS hostname or IP address, followed optionally by a : and a port. If the port is not specified, the default of 4556 is used. |
type | Link type (see note) | ALLWAYSON, ONDEMAND, SCHEDULED, OPPORTUNISTIC | Specify the link type |
conv_layer | udp, tcp, bt, file, eth, norm, null, extcl, serial | N/A | Specify a Convergence Layer Adapter (CLA) |
opt | Possible settings | Default | Comments |
mtu | number (MTU) | Set by link type | The MTU of the link, used to control proactive fragmentation. |
min_retry_interval | number (seconds) | Set by link type | The minimum amount to wait between attempts to re-open the link. Default is set by the various link types but can be overridden by configuration parameters. |
max_retry_interval | number (seconds) | Set by link type | The minimum amount to wait between attempts to re-open the link. Default is set by the various link types but can be overridden by configuration parameters. |
idle_close_time | number (seconds) | 30 for ONDEMAND or 0 for OPPORTUNISTIC & ALLWAYSON | Idle time before the link is closed. MUST be zero for always on links (i.e. they are never closed). Default is 30 seconds for on demand links, zero for opportunistic links. |
potential_downtime | number (seconds) | 30 | A conservative estimate of the maximum amount of time that the link may be down during 'normal' operation. Used by routing algorithms to determine how long to leave bundles queued on the down link before rerouting them. |
prevhop_hdr | true or false | false | Indicates whether or not to send the previous hop header on this link. |
cost | number (abstract cost) | 100 | The abstract cost of the link, used by routing algorithms. |
qlimit_bundles_high | number (bundles) | tbc | The high limits on the number of bundles that should be queued on the link. The high limits are used to indicate whether or not more bundles can be queued onto the link to effect backpressure from the convergence layers. |
qlimit_bundles_low | number (bundles) | tbc | The low limits on the number of bundles that should be queued on the link. The low limits can be used by the router to determine when to re-scan the pending bundle lists. |
qlimit_bytes_high | number (bytes) | tbc | The high limits on the number of bytes that should be queued on the link. The high limits are used to indicate whether or not more bundles can be queued onto the link to effect backpressure from the convergence layers. |
qlimit_bytes_low | number (bytes) | tbc | The low limits on the number of bytes that should be queued on the link. The low limits can be used by the router to determine when to re-scan the pending bundle lists. |
retry_interval | number (seconds) | min_retry_interval | The seconds to wait between attempts to re-open an unavailable link. Initially set to min_retry_interval, then doubles up to max_retry_interval |
Links are of types as discussed in the DTN architecture ALWAYSON, ONDEMAND, SCHEDULED, OPPORTUNISTIC. The key differences from an implementation perspective are "who" and "when" manipulates the link state regarding availability.
ONDEMAND links are initializd in the AVAILABLE state, as one would expect. It remains in this state until a router explicitly opens it. An ONDEMAND link can then be closed either due to connection failure or because the link has been idle for too long, both triggered by the convergence layer. If an ONDEMAND link is closed due to connection failure, then the contact manager is notified of this event and periodically tries to re-establish the link.
For OPPORTUNISTIC links the availability state is set by the code which detects that there is a new link available to be used. SCHEDULED links have their availability dictated by the schedule implementation.
All links in the OPEN state have an associated contact that represents an actual connection. Every link has a unique name associated with it which is used to identify it. The name is configured explicitly when the link is created.
The link dump command shows the following attributes for all links:
The link dump name command shows the following detailed attributes for a specified link:
The link stats name command shows the following counts for a specified link:
The test command is used by developers to test the DTN2 daemon. It is capable of generating a segfault, triggering a panic or false assert.
Syntax: test cmd
Example: test segfault