Opened 9 years ago

Closed 8 years ago

Last modified 8 years ago

#1778 closed task (implemented)

Regroup bw auth pairs and add a 5th one

Reported by: ln5 Owned by: mikeperry
Priority: High Milestone: Deliverable-Nov2011
Component: Core Tor/Torflow Version:
Severity: Keywords: MikePerryIteration20111106
Cc: ioerror, arma, phobos, karsten, linus@…, aagbsn Actual Points: 2
Parent ID: Points: 2
Reviewer: Sponsor:

Description

The connections between bw auths and dir auths look something like this atm:

arma -> arma
ln5 -> mike
mike -> ioerror (?)
phobos -> ?

This should be rewired. Please update the list above, those of you who know.

Child Tickets

Change History (25)

comment:1 Changed 9 years ago by mikeperry

Cc: karsten added

We also have:

karsten -> karsten.

I am thinking we should change to:
ln5 -> ln5
mike -> mike
phobos -> ioerror

comment:2 Changed 9 years ago by ln5

Cc: linus@… added

comment:3 Changed 9 years ago by mikeperry

Component: - Select a componentTorflow

comment:4 Changed 8 years ago by mikeperry

The current assignments are:
phobos -> moria1 (roger)
ln5 -> ides (mike)
bwscan vm -> urrass (ioerror)
karsten -> gabelmoo (karsten)

The minimal edit that produces something like sanity is:
phobos -> moria1 (roger)
ln5 -> maatuska (ln5)
bwscan vm -> urrass (ioerror)
karsten -> gabelmoo (karsten)
new scanner -> ides (mike)

Another possible alternative that doesn't require new servers is:
roger -> moria1 (roger)
ln5 -> maatuska (ln5)
bwscan vm -> urrass (ioerror)
karsten -> gabelmoo (karsten)
phobos -> ides (mike)

This depends on how well we expect moria1 to work with the postgres support.

comment:5 Changed 8 years ago by mikeperry

Cc: aagbsn added

Roger: Do you want to try out the new postgres support on moria? It is in master. We can coordinate setup on IRC if you like.

comment:6 Changed 8 years ago by mikeperry

Milestone: Deliverable-Nov2011

comment:7 Changed 8 years ago by mikeperry

Priority: minormajor

comment:8 Changed 8 years ago by karsten

In order to speed up things, shall I try out the new postgres support on gabelmoo? I already have PostgreSQL 8.3 running there for the auto-naming script. Or are we concerned that this breaks immediately?

comment:9 in reply to:  8 Changed 8 years ago by aagbsn

Replying to karsten:

In order to speed up things, shall I try out the new postgres support on gabelmoo? I already have PostgreSQL 8.3 running there for the auto-naming script. Or are we concerned that this breaks immediately?

Hold off for now. Roger has had some issues (#3834) and there are also a few commits I would like to get merged before proceeding.

comment:10 Changed 8 years ago by aagbsn

Commits are merged, and bwscan vm is now running the latest with postgres backend. (Thanks mikeperry!)

comment:11 Changed 8 years ago by aagbsn

bump.
karsten: go ahead with postgres support on gabelmoo.
Is anything else blocking?

comment:12 in reply to:  11 Changed 8 years ago by mikeperry

Replying to aagbsn:

bump.
karsten: go ahead with postgres support on gabelmoo.
Is anything else blocking?

Ok, Roger is OK with using moria for now. I say this means we do this:

roger -> moria1 (roger)
ln5 -> maatuska (ln5)
bwscan vm -> urrass (ioerror)
karsten -> gabelmoo (karsten)
phobos -> ides (mike)

I believe we can perform this change in three atomic operations:

  1. ln5 doesn't need to coordinate with any of us. He can make the change by himself (though he is of course welcome to ask me for help).
  1. ioerror must talk with me.
  1. phobos must talk with me.

comment:13 Changed 8 years ago by karsten

How's this going? We promised to set up a fifth scanner until June 30. Is this only blocking on people talking to each other, or are there things left to code and test, too?

comment:14 Changed 8 years ago by mikeperry

ln5 has committed to working with me at the end of Oct to reconnect salsa to maatuska. Then we should have 5, assuming moria1 can stay alive..

comment:15 Changed 8 years ago by ln5

/me nods.

comment:16 in reply to:  14 Changed 8 years ago by karsten

Replying to mikeperry:

ln5 has committed to working with me at the end of Oct to reconnect salsa to maatuska. Then we should have 5, assuming moria1 can stay alive..

It's almost end of Oct, and we need a fifth bwscanner running by the very first day of Nov. When did you plan to reconnect salsa to maatuska? And how's moria1 doing with the bwscanner?

comment:17 Changed 8 years ago by mikeperry

Keywords: MikePerryIteration20111106 added
Points: 2

comment:18 Changed 8 years ago by ln5

I'm unable to do the actual connecting before Mon 31st.
I would like to prepare as much as possible on salsa prior to that though.
I plan on doing this Fri 28th, evening (EST).

comment:19 Changed 8 years ago by ln5

My bw auth has been running for some 36h now and has measured 55% of the relays. Connecting it to maatuska within a couple of hours.

comment:20 Changed 8 years ago by ln5

Scanner #1 still hasn't managed to finish its slice, i.e. there are no done-files in the data/scanner.1/scan-data directory.

Scanners 2, 3 and 4 seem fine.

scanner.1/bw.log doesn't look much different than scanner.2/bw.log to me.

If this is reasonable I'll just wait some more. If there's something
I can do in order to produce some data for my dir auth before (some)
midnight that'd be great.

comment:21 Changed 8 years ago by ln5

s/finish its slice/finish enough relays/1

comment:22 Changed 8 years ago by ln5

So scanner #1 is still not producing. Mike and I have tried some various fun stuff in order to make it finish, such as

  • reducing the size of the slice for scanner 1 from 12 to 6 percent
  • lowering the timeout value from half an hour to five minutes
  • reducing the size of the fetched file from 8M to 4M for the top 5% and from 4M to 2M for the next 5%

Too early to tell if it works. Still none of the scanners has finished.

comment:23 Changed 8 years ago by ln5

maatuska will be voting for bandwidth the next time around (09:00 UTC).

comment:24 Changed 8 years ago by karsten

Resolution: implemented
Status: newclosed

And maatuska is indeed including [bandwidth scanner results in its vote! Great! Closing this ticket.

comment:25 Changed 8 years ago by mikeperry

Actual Points: 2
Summary: Bw authorities could report to dir authorities "closer" to themRegroup bw auth pairs and add a 5th one
Note: See TracTickets for help on using tickets.