mbox series

[net-next,v6,00/10] net: openvswitch: Add sample multicasting.

Message ID 20240628110559.3893562-1-amorenoz@redhat.com (mailing list archive)
Headers show
Series net: openvswitch: Add sample multicasting. | expand

Message

Adrián Moreno June 28, 2024, 11:05 a.m. UTC
** Background **
Currently, OVS supports several packet sampling mechanisms (sFlow,
per-bridge IPFIX, per-flow IPFIX). These end up being translated into a
userspace action that needs to be handled by ovs-vswitchd's handler
threads only to be forwarded to some third party application that
will somehow process the sample and provide observability on the
datapath.

A particularly interesting use-case is controller-driven
per-flow IPFIX sampling where the OpenFlow controller can add metadata
to samples (via two 32bit integers) and this metadata is then available
to the sample-collecting system for correlation.

** Problem **
The fact that sampled traffic share netlink sockets and handler thread
time with upcalls, apart from being a performance bottleneck in the
sample extraction itself, can severely compromise the datapath,
yielding this solution unfit for highly loaded production systems.

Users are left with little options other than guessing what sampling
rate will be OK for their traffic pattern and system load and dealing
with the lost accuracy.

Looking at available infrastructure, an obvious candidated would be
to use psample. However, it's current state does not help with the
use-case at stake because sampled packets do not contain user-defined
metadata.

** Proposal **
This series is an attempt to fix this situation by extending the
existing psample infrastructure to carry a variable length
user-defined cookie.

The main existing user of psample is tc's act_sample. It is also
extended to forward the action's cookie to psample.

Finally, a new OVS action (OVS_SAMPLE_ATTR_EMIT_SAMPLE) is created.
It accepts a group and an optional cookie and uses psample to
multicast the packet and the metadata.

--
v5 -> v6:
- Renamed emit_sample -> psample
- Addressed unused variable and conditionally compilation of function.

v4 -> v5:
- Rebased.
- Removed lefover enum value and wrapped some long lines in selftests.

v3 -> v4:
- Rebased.
- Addressed Jakub's comment on private and unused nla attributes.

v2 -> v3:
- Addressed comments from Simon, Aaron and Ilya.
- Dropped probability propagation in nested sample actions.
- Dropped patch v2's 7/9 in favor of a userspace implementation and
consume skb if emit_sample is the last action, same as we do with
userspace.
- Split ovs-dpctl.py features in independent patches.

v1 -> v2:
- Create a new action ("emit_sample") rather than reuse existing
  "sample" one.
- Add probability semantics to psample's sampling rate.
- Store sampling probability in skb's cb area and use it in emit_sample.
- Test combining "emit_sample" with "trunc"
- Drop group_id filtering and tracepoint in psample.

rfc_v2 -> v1:
- Accommodate Ilya's comments.
- Split OVS's attribute in two attributes and simplify internal
handling of psample arguments.
- Extend psample and tc with a user-defined cookie.
- Add a tracepoint to psample to facilitate troubleshooting.

rfc_v1 -> rfc_v2:
- Use psample instead of a new OVS-only multicast group.
- Extend psample and tc with a user-defined cookie.


Adrian Moreno (10):
  net: psample: add user cookie
  net: sched: act_sample: add action cookie to sample
  net: psample: skip packet copy if no listeners
  net: psample: allow using rate as probability
  net: openvswitch: add psample action
  net: openvswitch: store sampling probability in cb.
  selftests: openvswitch: add psample action
  selftests: openvswitch: add userspace parsing
  selftests: openvswitch: parse trunc action
  selftests: openvswitch: add psample test

 Documentation/netlink/specs/ovs_flow.yaml     |  17 ++
 include/net/psample.h                         |   5 +-
 include/uapi/linux/openvswitch.h              |  31 +-
 include/uapi/linux/psample.h                  |  11 +-
 net/openvswitch/Kconfig                       |   1 +
 net/openvswitch/actions.c                     |  65 ++++-
 net/openvswitch/datapath.h                    |   3 +
 net/openvswitch/flow_netlink.c                |  32 ++-
 net/openvswitch/vport.c                       |   1 +
 net/psample/psample.c                         |  16 +-
 net/sched/act_sample.c                        |  12 +
 .../selftests/net/openvswitch/openvswitch.sh  | 115 +++++++-
 .../selftests/net/openvswitch/ovs-dpctl.py    | 272 +++++++++++++++++-
 13 files changed, 565 insertions(+), 16 deletions(-)

Comments

Adrián Moreno June 28, 2024, 5:38 p.m. UTC | #1
On Fri, Jun 28, 2024 at 01:05:36PM GMT, Adrian Moreno wrote:
> ** Background **
> Currently, OVS supports several packet sampling mechanisms (sFlow,
> per-bridge IPFIX, per-flow IPFIX). These end up being translated into a
> userspace action that needs to be handled by ovs-vswitchd's handler
> threads only to be forwarded to some third party application that
> will somehow process the sample and provide observability on the
> datapath.
>
> A particularly interesting use-case is controller-driven
> per-flow IPFIX sampling where the OpenFlow controller can add metadata
> to samples (via two 32bit integers) and this metadata is then available
> to the sample-collecting system for correlation.
>
> ** Problem **
> The fact that sampled traffic share netlink sockets and handler thread
> time with upcalls, apart from being a performance bottleneck in the
> sample extraction itself, can severely compromise the datapath,
> yielding this solution unfit for highly loaded production systems.
>
> Users are left with little options other than guessing what sampling
> rate will be OK for their traffic pattern and system load and dealing
> with the lost accuracy.
>
> Looking at available infrastructure, an obvious candidated would be
> to use psample. However, it's current state does not help with the
> use-case at stake because sampled packets do not contain user-defined
> metadata.
>
> ** Proposal **
> This series is an attempt to fix this situation by extending the
> existing psample infrastructure to carry a variable length
> user-defined cookie.
>
> The main existing user of psample is tc's act_sample. It is also
> extended to forward the action's cookie to psample.
>
> Finally, a new OVS action (OVS_SAMPLE_ATTR_EMIT_SAMPLE) is created.
> It accepts a group and an optional cookie and uses psample to
> multicast the packet and the metadata.
>
> --
> v5 -> v6:
> - Renamed emit_sample -> psample
> - Addressed unused variable and conditionally compilation of function.
>
> v4 -> v5:
> - Rebased.
> - Removed lefover enum value and wrapped some long lines in selftests.
>
> v3 -> v4:
> - Rebased.
> - Addressed Jakub's comment on private and unused nla attributes.
>
> v2 -> v3:
> - Addressed comments from Simon, Aaron and Ilya.
> - Dropped probability propagation in nested sample actions.
> - Dropped patch v2's 7/9 in favor of a userspace implementation and
> consume skb if emit_sample is the last action, same as we do with
> userspace.
> - Split ovs-dpctl.py features in independent patches.
>
> v1 -> v2:
> - Create a new action ("emit_sample") rather than reuse existing
>   "sample" one.
> - Add probability semantics to psample's sampling rate.
> - Store sampling probability in skb's cb area and use it in emit_sample.
> - Test combining "emit_sample" with "trunc"
> - Drop group_id filtering and tracepoint in psample.
>
> rfc_v2 -> v1:
> - Accommodate Ilya's comments.
> - Split OVS's attribute in two attributes and simplify internal
> handling of psample arguments.
> - Extend psample and tc with a user-defined cookie.
> - Add a tracepoint to psample to facilitate troubleshooting.
>
> rfc_v1 -> rfc_v2:
> - Use psample instead of a new OVS-only multicast group.
> - Extend psample and tc with a user-defined cookie.
>
>
> Adrian Moreno (10):
>   net: psample: add user cookie
>   net: sched: act_sample: add action cookie to sample
>   net: psample: skip packet copy if no listeners
>   net: psample: allow using rate as probability
>   net: openvswitch: add psample action
>   net: openvswitch: store sampling probability in cb.
>   selftests: openvswitch: add psample action
>   selftests: openvswitch: add userspace parsing
>   selftests: openvswitch: parse trunc action
>   selftests: openvswitch: add psample test
>
>  Documentation/netlink/specs/ovs_flow.yaml     |  17 ++
>  include/net/psample.h                         |   5 +-
>  include/uapi/linux/openvswitch.h              |  31 +-
>  include/uapi/linux/psample.h                  |  11 +-
>  net/openvswitch/Kconfig                       |   1 +
>  net/openvswitch/actions.c                     |  65 ++++-
>  net/openvswitch/datapath.h                    |   3 +
>  net/openvswitch/flow_netlink.c                |  32 ++-
>  net/openvswitch/vport.c                       |   1 +
>  net/psample/psample.c                         |  16 +-
>  net/sched/act_sample.c                        |  12 +
>  .../selftests/net/openvswitch/openvswitch.sh  | 115 +++++++-
>  .../selftests/net/openvswitch/ovs-dpctl.py    | 272 +++++++++++++++++-
>  13 files changed, 565 insertions(+), 16 deletions(-)
>
> --
> 2.45.2
>

Patchwork says this patch is not applying on net-next. I'll wait for
some reviews and rebase+resubmit it later tonight or tomorrow.

Thanks.
Adrián