- install cds balancer by importing package cdsbalancer
- `net.SplitHostPort` fails when target doesn't have port, but we want to use the original string instead
This PR also adds a new testutils directory with a fake client and a
channel which supports a timed receive operation. In follow up PRs, we
can move other common test stuff like the fake server etc to this
directory and cleanup more tests.
The client will look only at the last route in the list (the default
route), whose match field must contain a prefix field whose value is the
empty string and whose route field must be set.
Continuing the war on stacks, we can reduce the amount of stack required
per-RPC by combining defers from different components into one.
Each defer statement in process{Unary,Streaming}RPC goes on the stack
and occupies about 56-64 bytes the entire lifetime of an RPC, which
could be very long. More importantly, a call to runtime.morestack is
often required to allocate a new, larger stack when the handler
goroutine runs out of stack memory (Go's default stack size is 2 KiB).
Before:
$ go tool objdump <binary> | grep "TEXT.*processUnaryRPC(SB)" -A 10 | grep "SUBQ.*SP"
server.go:867 0x9132fb 4881ec80030000 SUBQ $0x380, SP
$ go tool objdump <binary> | grep "TEXT.*processStreamingRPC(SB)" -A 10 | grep "SUBQ.*SP"
server.go:1099 0x9151bb 4881ec68020000 SUBQ $0x268, SP
After:
$ go tool objdump <binary> | grep "TEXT.*processUnaryRPC(SB)" -A 10 | grep "SUBQ.*SP"
server.go:867 0x9132fb 4881ecd0020000 SUBQ $0x2d0, SP
$ go tool objdump <binary> | grep "TEXT.*processStreamingRPC(SB)" -A 10 | grep "SUBQ.*SP"
server.go:1116 0x9150fb 4881ecf8010000 SUBQ $0x1f8, SP
As one can observe, the processUnaryRPC's stack goes down from 0x380
bytes to 0x2d0 bytes (896 - 720 = 176 bytes) while processStreamingRPC's
stack goes down from 0x2d8 bytes to 0x1f8 bytes (616 - 504 = 112 bytes).
There are probably other things we can do here, but these are some low
hanging fruits to pick off.
This PR removes the xds_client implementation from eds balancer, and replaces it with a xds_client wrapper. (The xds_client wrapper has very similar API as the old xds_client implementation, so the change in the eds balancer is minimal).
The eds balancer currently doesn't look for xds_client from attributes, and always creates a new xds_client. The attributes change will be done in a following up change.
* Add a helper to the fakexds package to return a ClientConn talking to
the fake server.
* Tests will make use of this ClientConn wherever required, or they will
directly pass the fake server's address as the balancerName to the
xdsclient.New() function, thus exercising that code path as well.
* Add grpc.WithTimeout to list in vet.sh
The xds client will parse the EDS response, and give the parse result to eds balancer, so the balancer doesn't need to deal with proto directly.
Also moved `ClusterLoadAssignmentBuilder` to another pacakge to be shared by tests in different packages.
Generated protobuf messages contain internal data structures
that general purpose comparison functions (e.g., reflect.DeepEqual,
pretty.Compare, etc) do not properly compare. It is already the case
today that these functions may report a difference when two messages
are actually semantically equivalent.
Fix all usages by either calling proto.Equal directly if
the top-level types are themselves proto.Message, or by calling
cmp.Equal with the cmp.Comparer(proto.Equal) option specified.
This option teaches cmp to use proto.Equal anytime it encounters
proto.Message types.
Each priority maps to a balancer group.
When a priority is in use, its balancer group is started, and it will close the balancer groups with lower priorities. When a priority is down (no connection ready), it will start the next priority balancer group.
