diff options
author | Mark Linimon <linimon@FreeBSD.org> | 2010-06-22 23:48:07 +0000 |
---|---|---|
committer | Mark Linimon <linimon@FreeBSD.org> | 2010-06-22 23:48:07 +0000 |
commit | e8f9e57ecaa9dd75b0463a28e9dd468ce0d43c93 (patch) | |
tree | e83ef9721e0a3e78888eedc9c9995611e5822565 /Tools/portbuild | |
parent | 408042ec04de3b5366a4753a6cf8ec57579c503c (diff) |
Notes
Diffstat (limited to 'Tools/portbuild')
-rw-r--r-- | Tools/portbuild/conf/README.dotunnel | 32 |
1 files changed, 32 insertions, 0 deletions
diff --git a/Tools/portbuild/conf/README.dotunnel b/Tools/portbuild/conf/README.dotunnel new file mode 100644 index 000000000000..b3a4fd5a12ab --- /dev/null +++ b/Tools/portbuild/conf/README.dotunnel @@ -0,0 +1,32 @@ +Various package build nodes require us to set up TCP tunnels to talk +to them. (Some systems don't pass certain ports; some systems have +firewalls; some systems have multiple nodes on one IP address.) + +These have always been hardcoded in crontab lines of the form "while true; +do <hardcoded-tunnel-stuff>; done". Other than the magic hardcoding, +there's a problem with this. When the tunnel command exits, such as if +the host suddenly becoming unreachable, it doesn't send mail -- instead +it just accumulates a huge file in /var/spool/clientmqueue which never +gets sent. To add insult to injury, /var is on the root partition on +pointyhat. + +To cure these problems, we now have + + /var/portbuild/conf/<arch>/dotunnel.XXX + +where XXX corresponds to one line in the old crontab. Each script sets +up one tunnel, sends mail to the user(s) in + + /var/portbuild/<arch>/portbuild.conf + +once the command exits, and then sleeps. + +Why not put it in /var/portbuild/<arch> you ask? That directory is +propogated to all nodes for that arch. This would be a security leak. +The intention is that none of the dotunnel files will be checked into +CVS. + +Final note: each script figures out which arch it is for by fiddling +with its $0, so invoke it with its full pathname. + +mcl |