RspRegistrar(1)

📦
version 3.5.8 · 4 versions
, in ubuntu stonking

💖 RspRegistrar(1) — the RSerPool Pool Registrar (PR) service, bestie edition 💖

okay so you wanna run a Pool Registrar? that's rspregistrar, from the General Commands Manual, and she is iconic. sit down, grab a strawberry 🍓, and let's get into it.

✨ SYNOPSIS ✨

no cap, this is the whole vibe:

rspregistrar [-announcettl=TTL] [-asap=auto|address:port,address,...]
             [-asapannounce=auto|address:port] [-autoclosetimeout=seconds]
             [-endpointkeepalivetransmissioninterval=milliseconds]
             [-endpointkeepalivetimeoutinterval=milliseconds]
             [-maxbadpereports=reports] [-maxhresitems=items]
             [-maxincrement=increment]
             [-minaddressscope=loopback|sitelocal|global]
             [-serverannouncecycle=milliseconds]
             [-enrp=auto|address:port,address,...]
             [-enrpannounce=auto|address:port]
             [-maxelementsperhtrequest=items]
             [-mentordiscoverytimeout=milliseconds] [-peer=address:port]
             [-peerheartbeatcycle=milliseconds]
             [-peermaxtimelastheard=millisecond]
             [-peermaxtimenoresponse=milliseconds]
             [-takeoverexpiryinterval=milliseconds]
             [-cspinterval=milliseconds] [-cspserver=address:port]
             [-logcolor=on|off] [-logappend=filename] [-logfile=filename]
             [-loglevel=0-9]

she's serving many flags and honestly? she ate 💅

🌸 DESCRIPTION 🌸

okay so rspregistrar runs a registrar service for Reliable Server Pooling (RSerPool) pool elements (PEs) and pool users (PUs). the PRs inside one operation scope keep their view of the handlespace in sync using the Endpoint haNdlespace Redundancy Protocol (ENRP), and that whole thing rides on SCTP, obviously.

so an operation scope has a limited range — like, an organization, or literally just one building, that kind of thing. and here's the important part: it's restricted to a single administrative domain. that's the whole contrast with GRID computing, and the reason is keeping management complexity at a minimum, fr fr. but like — the PEs themselves can still be distributed globally so that the service survives localized disasters. slay, resilience. 🦋

now, PEs pick an arbitrary PR of the operation scope to register into a pool, and they do it with the Aggregate Server Access Protocol (ASAP) — again over SCTP, because of course. the moment a PE registers at a PR, that PR becomes the Home-PR (PR-H) of the newly registered PE, she's the main character now 👑.

and a PR-H is responsible for monitoring its PEs' availability with keep-alive messages — which the PE has to acknowledge within a given timeout, no ghosting allowed 😭 — and then it propagates the info about its PEs to the other PRs of the operation scope via ENRP updates.

so then how does a PU get to a pool? a PU requests a PE selection from an arbitrary PR of the operation scope, using ASAP, transported via SCTP. and the PR picks the requested list of PE identities by applying a pool-specific selection rule, which is called the pool policy. it's giving matchmaker ✨

and because we love automatic configuration, a PR can send out regular announces via an ASAP and ENRP multicast address. PEs and PUs just listen to that address to learn the list of currently available PRs. no manual config, just vibes. 🎀

🎀 ARGUMENTS 🎀

the following options are available:

  • -quiet — don't print startup and shutdown messages. silence is golden, bestie. 🤫
  • -announcettl=TTL — sets the TTL for outgoing ASAP Announce/ENRP Presence messages via multicast.

📝 Logging Parameters

quick heads up: it is recommended to use at least a value of 2 so you can actually see possibly important error messages and warnings. don't be shy with the logging.

  • -logcolor=on|off — turns ANSI colorization of the logging output on or off.
  • -logappend=filename — write logging output to a file (default is stdout); output will be appended to the given file.
  • -logfile=filename — write logging output to a file (default is stdout); the given file will be overwritten.
  • -loglevel=0-9 — sets the logging verbosity from 0 (none) to 9 (very verbose).

🩺 Component Status Protocol (CSP) Parameters

  • -cspinterval=milliseconds — sets the interval for the CSP status updates in milliseconds.
  • -cspserver=address:port — sets the CSP monitor server's address and port.

🌈 Aggregate Server Access Protocol (ASAP) Parameters

  • -autoclosetimeout=seconds — sets the SCTP autoclose timeout for idle ASAP associations.
  • -asap=auto|address:port,address,... — sets the ASAP endpoint address(es). use "auto" to automatically set it (default).
  • -asapannounce=auto|address:port — sets the multicast address and UDP port to send the ASAP Announces to. use "auto" to automatically set it (default). examples:
-asapannounce=auto
-asapannounce=239.0.0.1:3863
  • -endpointkeepalivetransmissioninterval=milliseconds — sets the ASAP Endpoint Keep Alive interval.
  • -endpointkeepalivetimeoutinterval=milliseconds — sets the ASAP Endpoint Keep Alive timeout.
  • -maxbadpereports=reports — sets the maximum number of ASAP Endpoint Unreachable reports before removing a PE.
  • -maxincrement=increment — sets the MaxIncrement constant. ⚠️ Handle with care!
  • -maxhresitems=items — sets the MaxHResItems constant.
  • -minaddressscope=loopback|sitelocal|global — sets the minimum address scope acceptable for registered PEs:
    • loopback: Loopback address (only valid on the same node!)
    • site-local: Site-local addresses (e.g. 192.168.1.1, etc.)
    • global: Global addresses
  • -serverannouncecycle=milliseconds — sets the ASAP Announce interval.

🦋 Endpoint Handlespace Redundancy Protocol (ENRP) Parameters

  • -enrp=auto|address:port,address,... — sets the ENRP endpoint address(es). use "auto" to automatically set it (default). examples:
-enrp=auto
-enrp=1.2.3.4:9901
-enrp=1.2.3.4:9901,[2000::1:2:3],9.8.7.6
  • -enrpannounce=auto|address:port — sets the multicast address and UDP port to send the ENRP Announces to. use "auto" to automatically set it (default). examples:
-enrpannounce=auto
-enrpannounce=239.0.0.1:9901
  • -maxelementsperhtrequest=items — sets the maximum number of items per ENRP Handle Table Response.
  • -mentordiscoverytimeout=milliseconds — sets the mentor PR discovery timeout in milliseconds.
  • -peer=address:port — adds a static PR entry into the Peer List. it is possible to add multiple entries. 👯
  • -peerheartbeatcycle=milliseconds — sets the ENRP peer heartbeat interval.
  • -peermaxtimelastheard=milliseconds — sets the ENRP peer max time last heard.
  • -peermaxtimenoresponse=milliseconds — sets the ENRP maximum time without response.
  • -takeoverexpiryinterval=milliseconds — sets the ENRP takeover timeout.

🌱 ENVIRONMENT 🌱

rspregistrar uses the environment variables CSP_SERVER and CSP_INTERVAL to define a CSP server to send reports to in the specified interval. yes ma'am, env vars came through 💜

👀 DIAGNOSTICS 👀

if loglevel>0, log messages will be printed to stdout or into a specified log file.

🔗 SEE ALSO 🔗

for a detailed introduction to RSerPool, see: https://duepublico.uni-duisburg-essen.de/servlets/DerivateServlet/Derivate-16326/Dre2006_final.pdf

Thomas Dreibholz's RSerPool Page: https://www.nntb.no/~dreibh/rserpool/

and that's the whole page, bestie — RSerPool Registrar, May 12, 2025. go be the main character of your handlespace 🫶✨

📦 where this page comes from