| Commit message (Collapse) | Author | Age |
|---|
| ... | |
| |
|
|
| |
Includes suggestions by tomlurge.
|
| |
|
|
| |
Implements #17321.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
| |
Suggested by atagar. Descriptors of this @type are not archive by
CollecTor, but libraries like Stem that support parsing many of the
descriptors listed on CollecTor also support parsing this type, so it
makes sense to list it here.
Implements #15009.
|
| | |
|
| | |
|
| |
|
|
|
|
| |
We're not providing these files via HTTP, and we stopped rsync'ing
them locally when moving ExoneraTor to another host. No need to waste
all those fine inodes.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
Spotted by Nikita Borisov.
|
| |
|
|
| |
Implements #12676.
|
| |
|
|
|
| |
Kostas informs me that the service has become unavailable, so let's take
out the link for now.
|
| |
|
|
|
|
| |
These lines are being added as part of #12538, and even though this
code has not been merged to master, we shouldn't discard descriptors
containing them. The exact format might change in the future.
|
| | |
|
| | |
|
| |
|
|
| |
Spotted by phw.
|
| |
|
|
|
|
| |
Related to #12951 that adds this line. Before that change, we'd simply
use the file's last-modified time as publication time and write that to
the sanitized status file.
|
| |
|
|
|
|
|
| |
turtles doesn't participate in the consensus for a couple of days now, and
it times out on our connections. There's no way how we can fetch all
descriptors referenced from its vote, so there's a new warning every hour.
Ignore.
|
| |
|
|
| |
Spotted by cypherpunks, resolves #13265.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A few days ago there was an unusual situation with parsing gabelmoo's
cached descriptors and fetching descriptors from the other directory
authorities. gabelmoo failed to produce a consensus at :00 of the hour
and downloaded the consensus from the other directory authorities after
:05 of the hour. As a result, gabelmoo's cached consensus was 1 hour
older than the consensus that we downloaded from the other directory
authorities.
Our stale check considers all stored descriptors and determines if there
were any consensuses for which we didn't also store votes or microdesc
consensuses. In this case the stale check triggered, because we didn't
store any votes or microdesc consensuses with the same valid-after time as
gabelmoo's outdated cached consensus. This warning was confusing.
The fix is simple: if we're storing a descriptor that we stored before,
don't include it in the stale check.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
| |
In b65f0a1, we included ntor-onion-key lines in sanitized bridge server
descriptors but raised the @type version number of sanitized bridge
extra-infos. Too late to fix that, but let's document this.
Spotted by Isis.
|
| |
|
|
|
| |
At this point, everything relevant from the metrics website should be
included.
|
| |
|
|
|
| |
From now on we're going to provide Tor network data ourselves, rather than
relying on metrics-web to do that for us.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
We're currently providing server descriptors, extra-info descriptors, and
microdescriptors in one file per descriptor. The result is an rsync
recent directory containing 300,000 files, which takes forever to sync.
The new approach is to concatenate descriptors of the same type into one
file per execution. We're still deleting files after 72 hours, so that
there will always be three days of data available. But the same data only
occupies about 900 files, which is much more efficient to sync.
The downside is that rsync clients will have to learn how to split up
files into single descriptors. Let's give them some time to adapt.
|
| | |
|
| | |
|
| |
|
|
| |
These lines only contain exit policy summaries, not IP addresses.
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
Our internal descriptor completeness statistics just broke, because we
stopped remembering that we stored a microdesc more than 2 days ago. For
server descriptors and extra-info descriptors 2 days are sufficient,
because they won't be referenced in a consensus after that. But
microdescs may be used for a week or more. In fact, raise period to 40
days to be sure that we downloaded microdescs this month. This only
affects internal statistics, nothing else.
|
| |
|
|
|
|
| |
Also support reading microdescriptors from disk.
Implements #2785.
|
| |
|
|
| |
Added in #9859.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This ports commit 3e60ccdaaba598cabb7281d45f9a415299b8e3e3 from
metrics-lib over, originally written by Jens-Michael Hoffmann
<jmho@jmho.de>, with minor tweaks by Karsten Loesing.
The local lib directory is not used anymore and respective references
were removed. The java dependencies are now specified in the build.xml
and taken from their installed locations.
In addition to git, openjdk-6-jdk and ant the following java packages
have to be installed:
- libcommons-codec-java
- libcommons-compress-java
Slighty tweaked by Sebastian Hahn to remove references to unused junit.
|
| | |
|
| |
|
|
|
| |
Ian says in #8028 that doing so seems safe to him, because it's just a
random group element.
|
| |
|
|
|
| |
Instead, change cronjob to */6, which is possible now that metrics-db has
five separate modules.
|
| | |
|
