On Wed, May 27, 2009 at 05:10:07PM -0400, Brother Railgun of Reason wrote: > and then telnetting to port 3493, still from localhost, and I still > couldn't connect to port 3493. So, being a nasty suspicious sort, I > checked lsof ... nothing's got port 3493 open. checked the process > table ... upsd isn't running. It appears to start up, then immediately > die. > > Could this be because something is limiting it to 256 connections? > (Not that I'm ever going to need even a tenth of that.... 16 > connections would be more than I'm ever going to need to use.) > > I'm wondering whether I should go tweak the source to request only 256 > connections (or even 128) and see if upsd still dies on startup.
Indeed, I checked the code, and inability to get the requested number of connections is a fatal error. Investigating this, I've discovered that when upsd checks how many connections it may have, it's actually checking _SC_OPEN_MAX, which is the number of files it's allowed to have open, and which appears to be governed by rlim_fd_cur. On Solaris 10, out of the box, rlim_fd_cur defaults to 256. I'll reset this to 2048 in /etc/system on my next reboot, but in the meantime I've patched upsd.c to set maxconn to 128, since I'm never going to need even 256 connections to upsd. Now: babylon4:root:/opt/nut:10 # upsd Network UPS Tools upsd 2.4.1 listening on 10.24.32.14 port 3493 listening on 127.0.0.1 port 3493 Connected to UPS [tokamak]: bcmxcp-tokamak babylon4:root:/opt/nut:11 # upsc toka...@localhost ups.status OL babylon4:root:/opt/nut:12 # upsc toka...@localhost ambient.temperature: 23.8 ambient.temperature.high: 55 battery.charge: 33.1 battery.runtime: 442 battery.voltage: 129.2 driver.name: bcmxcp driver.parameter.baud_rate: 19200 driver.parameter.pollinterval: 2 driver.parameter.port: /dev/tty00 driver.version: 2.4.1 driver.version.internal: 0.21 input.current: 11.2 input.frequency: 60.0 input.frequency.high: 65 input.frequency.low: 55 input.frequency.nominal: 60 input.transfer.boost.high: 12 input.voltage: 119.8 input.voltage.nominal: 120 outlet.1.delay.shutdown: -1 outlet.1.delay.start: 1 outlet.1.id: 1 outlet.1.status: On outlet.2.delay.shutdown: -1 outlet.2.delay.start: 2 outlet.2.id: 2 outlet.2.status: On outlet.3.delay.shutdown: -1 outlet.3.delay.start: 3 outlet.3.id: 3 outlet.3.status: On output.current: 9.4 output.current.nominal: 24.0 output.frequency: 60.0 output.phases: 1 output.voltage: 115.3 output.voltage.nominal: 120 ups.beeper.status: enabled ups.firmware: Cont:01.01 ups.load: 37.5 ups.model: COMPAQ UPS 2900VA ups.power: 1080.5 ups.power.nominal: 2900 ups.serialMPAQ ups.status: OL So this is starting to look much better. I wish I'd realized sooner that upsd wasn't just complaining, but complaining *and dying*. -- Phil Stracchino, CDK#2 DoD#299792458 ICBM: 43.5607, -71.355 [email protected] [email protected] [email protected] Renaissance Man, Unix ronin, Perl hacker, Free Stater It's not the years, it's the mileage. _______________________________________________ Nut-upsuser mailing list [email protected] http://lists.alioth.debian.org/mailman/listinfo/nut-upsuser
