Age | Commit message (Collapse) | Author |
|
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Shrink memory usage.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
https://github.com/osrg/gobgp/issues/1249
The IN policy was removed. The modification by the IMPORT policy are
visible to all route server peers.
This saves some dozens bytes memory per a path.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
We don't need nanosecond precision. shrinks from 96 to 80 bytes.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
With Destination struct, multiple Paths can be moved from the table
package to the server package in a cleaner way; for features such as
add-path.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
The usage of "github.com/sirupsen/logrus" is recommended.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
This patch adds multiPathList field to watcherEventBestPathMsg and
fills it when global.use-multiple-paths.config.enable = true
Following patches add support injecting multipath to zebra and
monitoring it via gRPC
Signed-off-by: ISHIDA Wataru <ishida.wataru@lab.ntt.co.jp>
|
|
We need to update yang config collector mode (split into config and
state). before that, let's remove the feature itself to see if there
is any user of it.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
older path get preference
can be disabled by
`global.route-selection-options.external-compare-router-id = true`
close #806
Signed-off-by: ISHIDA Wataru <ishida.wataru@lab.ntt.co.jp>
|
|
Nobody uses them.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Later, we move non-bgp protocol stuff like mrt under their own
direcotries.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Withdrawn pathes are kept to be referenced thus the memory for them
are not freed. Kill oldKnownPathList and Destination's NewFeed().
Destination's Calculate() returns the best paths.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
validation
Withdrawn pathes are kept to be referenced thus the memory for them
are not freed. Nobody uses this so let's remove it.
Paths in UpdatedPathList are also in KnownPathList so doesn't lead to
memory leak. But remove it for simplicity.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
fix adjrib regression due to the d2801681 commit.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
the result of memory profile (500 route-server-clients each of them advertises
100 routes)
before:
(pprof) top5
9330.48MB of 9367.53MB total (99.60%)
Dropped 157 nodes (cum <= 46.84MB)
Showing top 10 nodes out of 17 (cum >= 9334.17MB)
flat flat% sum% cum cum%
6163.04MB 65.79% 65.79% 6163.04MB 65.79% github.com/osrg/gobgp/table.NewPath
1155.05MB 12.33% 78.12% 7302.59MB 77.96% github.com/osrg/gobgp/table.(*Path).Clone
986.31MB 10.53% 88.65% 1388.81MB 14.83% github.com/osrg/gobgp/table.(*AdjRib).Update
402.51MB 4.30% 92.95% 402.51MB 4.30% fmt.Sprintf
402.51MB 4.30% 97.24% 402.51MB 4.30% net.parseIPv4
after:
(pprof) top
3913.02MB of 3978.69MB total (98.35%)
Dropped 148 nodes (cum <= 19.89MB)
Showing top 10 nodes out of 11 (cum >= 21MB)
flat flat% sum% cum cum%
2970.30MB 74.66% 74.66% 2975.80MB 74.79% github.com/osrg/gobgp/server.filterpath
810.09MB 20.36% 95.02% 810.59MB 20.37% github.com/osrg/gobgp/table.(*AdjRib).Update
115.60MB 2.91% 97.92% 119.10MB 2.99% github.com/osrg/gobgp/table.createUpdateMsgFromPath
10MB 0.25% 98.17% 1878.02MB 47.20% github.com/osrg/gobgp/server.(*BgpServer).propagateUpdate
4.50MB 0.11% 98.29% 144.60MB 3.63% github.com/osrg/gobgp/table.CreateUpdateMsgFromPaths
Signed-off-by: ISHIDA Wataru <ishida.wataru@lab.ntt.co.jp>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
speed up and reduce memory footprint
Signed-off-by: ISHIDA Wataru <ishida.wataru@lab.ntt.co.jp>
|
|
Signed-off-by: ISHIDA Wataru <ishida.wataru@lab.ntt.co.jp>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Make the code simpler
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Signed-off-by: ISHIDA Wataru <ishida.wataru@lab.ntt.co.jp>
|
|
Signed-off-by: ISHIDA Wataru <ishida.wataru@lab.ntt.co.jp>
|
|
TableManager.localAsn wasn't used.
Signed-off-by: ISHIDA Wataru <ishida.wataru@lab.ntt.co.jp>
|
|
$ gobgp mrt dump 10 -o ./dump
Signed-off-by: ISHIDA Wataru <ishida.wataru@lab.ntt.co.jp>
|
|
Path/Destination/Table was defined as interface.
Simplify by making them normal struct.
Signed-off-by: ISHIDA Wataru <ishida.wataru@lab.ntt.co.jp>
|
|
This reverts commit f8832eb4811f57d16211fe31986b395ffb3c6ff0.
|
|
Signed-off-by: ISHIDA Wataru <ishida.wataru@lab.ntt.co.jp>
|
|
Signed-off-by: ISHIDA Wataru <ishida.wataru@lab.ntt.co.jp>
|
|
- add ToApiStruct() for convertion of internal structs to protobuf structs
to avoid ugly convertion by json.Marshal() && json.Unmarshal()
- move grpc server code under /server instead of /api
- update proto file to include more detailed path information
Signed-off-by: ISHIDA Wataru <ishida.wataru@lab.ntt.co.jp>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
struct Path has a flag which can tell whether the path is withdrawal
path or not. so table.ProcessPaths() needs not to return withdrawal path
and non-withdrawal path separetly.
this removes such unnecessary distinction.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
NewTableManager creates tables for only specified route families.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
|
|
We update rib-in first and then local. So should handle timestamp
before adding path to rib-in.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
If a new path has the exact same attributes that the existing one
(after RouteRefresh), we don't update the timestamp.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
must use Prefix (address + mask) as key.
Same as commit 900cb321fb3dc353b102aa1970e652184abc24eb
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
update unit tests.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Even if a connection with peer is down and up, table code gets
BGPmessages belonging the first connection before the event of the
peer down in order. So we don't need source version stuff inherent
from Ryu BGP code.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Always use 4byte for AS internally. Convert AS for 2byte AS peer when
necessary.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
table_manager.go clones withdraw path before passing it to peer.go
clone() doesn't clone pathattributes. This patch adds a helper
funciton to clone PathAttributes to message.go (it is used by
createing BGPUpdate message to withdraw ipv6. later it will be used
for 2 byte AS peer).
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
ProcessPaths and DeletePathsforPeer API uses both Path and Destination
for outside. There is no good reason to use Destination. Let's use
Path.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
RouteFamily is defined in two files.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
In some cases, we need to modify pathattributes in Path and creates an
update message based on it. Clone() is used to avoid chaning the
pathattributes in the table.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|