If multiple databases were specified, then the earlier writes would be
written to the later database and/or retention policy because points
were only written out after the batch size was reached. This forces the
batcher to flush whenever the database context switches.
This reverts commit 9aeae7ce82, reversing
changes made to 35b44cc2f0.
The contributor was unable to sign the contributor license agreement so
we have to revert this commit.
It inserted the changelog entry into version 1.0 because it couldn't
find any tags. Jenkins had been configured not to fetch any tags so the
changelog generator assumed this was the first version of the software.
Fetching tags is now enabled on Jenkins.
The default port for HTTP is 80 and HTTPS is 443. When you host InfluxDB
on a service that uses a reverse proxy, the `Host` header needs to be
exactly correct. Since the host header isn't expecting to see the port
when it is the default, this causes routing to fail.
This commit adds initial empty sketches back to the tsi1 index, as well
as ensuring that ephemeral sketches in the index `LogFile` are updated
accordingly.
The commit also adds a test that verifies that the merged sketches at
the store level produce the correct results under writes, deletions and
re-opening of the store.
This commit does not provide working sketches for post-compaction on the
tsi1 index.
The changelog is now updated automatically, tests are reported to github
as part of the CI status, and rebased/merged are also covered by
built-in github functionality.
On the master or release branches, `git changelog` will be run to
automatically update the changelog based off of the pull requests that
have been merged and then it will commit and push those changes.
Upstream must have changed and now autoconf isn't installed as an
implicit dependency of one of the packages anymore. Install is
explicitly so the fpm portion of the circle build works correctly.
batchWrite was using the last database and retention policy read from
the input file. Because batchWrite was called only every batchSize lines
or at EOF, databases with fewer than batchWrite points could be imported
into the incorrect database.
This change forces a flush with batchWrite whenever processDML reads a
change in database or retention policy.
Because of a race between the index and series file lookups, empty
keys can be returned for series which are tombstoned after the
series ids are obtained but before the caller looks up the key.
Remove some miscellaneous whitespace at the end of changelog entries and
fix two older entries. One accidentally included an inline html tag and
didn't display. The other was missing a newline between two changelog
entries and so didn't display correctly as separate bullet points.
The default of 4096 results in writes to the WAL still requiring muliple
IOs. We had previously bumped this to 1M, but that was too high when
there are many shards. Increasing to around 16k reduces the IOs to
one or two for the workloads tested. We may want to make this
configurable in the future.
Jonathan A. Sternberg
Edd Robinson
Stuart Carnie
Hercules Mango Churchill
Andrew Hare
Stuart Carnie
Ben Johnson
kun
Martin Geno
Walter Wilfinger
Adam
Jason Wilder
Jason Wilder