The Vector team is pleased to announce version 0.32.0!
Be sure to check out the upgrade guide for breaking changes in this release.
In addition to the usual enhancements and bug fixes, this release includes:
greptimedb
sink for sending metrics to
GreptimeDBprotobuf
codec that can be used on sources that support codecs to decode incoming
protobuf dataA number of sinks emit incorrect telemetry for the component_sent_*
metrics:
This is fixed in v0.32.1.
--openssl-legacy-provider
flag cannot actually be disabled by setting
it to false
via --openssl-legacy-provider=false
. Instead it complains of extra
arguments. This is fixed in v0.32.1.assume_role
for authentication without an external_id
caused a panic. This is fixed in v0.32.2`.clickhouse
sink database
and table
options are now
templatable.prometheus_scrape
source now scrapes configured targets in parallel.
Thanks to
nullren
for contributing this change!prometheus_scrape
source now has a scrape_timeout_secs
option to configure
how long Vector should wait for each request.
Thanks to
nullren
for contributing this change!debian
Docker images are now based on Debian 12 (Bookworm).protobuf
as an option. A Protobuf
descriptor file must also be provided to use to decode the data.
Thanks to
Daniel599
for contributing this change!VRL’s encrypt
and decrypt
functions now support additional algorithms:
CHACHA20-POLY1305
XCHACHA20-POLY1305
XSALSA20-POLY1305
AES-*-CTR-BE
(to disambiguate endianess of AES-*-CTR
)AES-*-CTR-LE
(to disambiguate endianess of AES-*-CTR
)nats
source and sink have been switched to use a more modern NATS library to
lay the groundwork for a JetStream source.
Thanks to
paolobarbolini
makarchuk
for contributing this change!greptimedb
sink was added allowing Vector to send metrics to
GreptimeDB.
Thanks to
sunng87
for contributing this change!log_schema.timestamp_key
) are
now parsed at boot-time rather than run-time. In addition to better performance,
this also means that invalid paths return an error at start time rather than being
silently ignored at runtime.lua
transform now sets the source_id
metadata to its own component ID if an
event is emitted by the transform that has no origin source_id
(e.g. events
constructed in the transform itself).filter
transform) are now
checked at boot-time to ensure that they return a boolean instead of treating all
non-boolean return values as false
.vector
sink now considers DataLoss
responses to be hard errors at indicates
the a sink in the downstream vector
source rejected the data. The vector
sink
will now not retry these errors and also reject them in any connected sources (when
acknowledgements
are enabled).
Thanks to
sbalmos
for contributing this change!vector
sink now correctly applies configured HTTPS proxy settings. Previously
it would fail to validate the downstream certificate.
Thanks to
joemiller
for contributing this change!splunk_hec
source now treats the fields on incoming events as “flat” rather
than interpreting them as field paths. For example, an incoming foo.bar
field is
now inserted as {"foo.bar": "..."}
rather than {"foo": {"bar": "..."}}
. This
avoids panics that were caused by invalid paths.api.enabled
to false
.
Thanks to
KH-Moogsoft
for contributing this change!component_received_event_bytes_total
and component_sent_event_bytes_total
metrics for sinks are now calculated after any encoding.only_fields
or
encoding.except_fields
options are applied.websocket
sink now correctly sends data as binary for “binary” codecs: raw
,
native
, and avro
. Previously it would always interpret the bytes as text
(UTF-8).
Thanks to
zhongchen
for contributing this change!syslog
source now correctly handles escape sequences appearing the structured
data segment.initial_concurrency
to start the concurrency limit at rather than starting at
a limit of 1
.
Thanks to
blake-mealey
for contributing this change!encode_logfmt
function now escapes all values including =
s.parse_nginx_log
function now handles more combined
formats.
Thanks to
scMarkus
for contributing this change!azure_blob_storage
sink now sets the correct content-type based on the
configured encoding
options.
Thanks to
stemjacobs
for contributing this change!vector
source no longer fails to decode large payloads. This was a regression
in 0.31.0 when a 4 MB limit was inadvertently applied.Sign up to receive emails on the latest Vector content and new releases
Thank you for joining our Updates Newsletter