Instead try pinging the client and let that decide whether the client
is alive or not. Ideally, it should always be failed since a well
behaved client would have closed the connection.
* This will allow client to continue monitoring peer connections while
connection stays open, so that there is no period of "uncertainity"
without previous no-monitoring approach.
* Use multiplier for incoming connection idle timeout
* Update idle connection timeout values and leaked connection timeout in
simulator.
This will not initiate request to get get new set of proxy unless we
know for a fact that endpoint has indeed failed, not just because the
connection to Peer was closed as it was sitting idle.
This patch does two changes to connection monitoring:
1. Connection monitoring at client side will check if the connection
has been stayed idle for some time. If connection is unused for a
while, we close the connection. There is some weirdness involved here
as ping messages are by themselves are connection traffic. We get over
this by making it two-phase process, first being checking idle
reliable traffic, followed by disabling pings and then checking for
idle unreliable traffic.
2. Connection monitoring of clients from server will no longer send
pings to clients. Instead, it keep monitor the received bytes and
close after certain period of inactivity.
RequestStream add another count to peerReference, which means as long
as ConnectionMonitor is alive, we'll never get peerReference=0 keeping
unnecessary connections potentially alive.
When simulation ends, all the actors are cancelled, and the
destructions which rely on `globals` may not have access to right
globals (instead of the default simulator process globals). This
patch, calls destroy on each process individually after we context
switch to that process so that the globals acceses in destructor are
its own.
This issue arised when trying to get `Peer::peerReferences` in
NetNotifiedQueue, resulting in decrementing the reference count of
peers in FlowTransport object of '0.0.0.0'.
The constructor of FlowReceiver which handled reference counting
peerReferences relied on calling a virtual method from constructor
whose behaviour isn't correct. This patch, bubbles down result of that
virtual method from derived constructor to base contructor.
Formerly, they would prefer to peek from the primary's logs. Testing of
a failed region rejoining the cluster revealed that this becomes quite a
strain on the primary logs when extremely large volumes of peek requests
are coming from the Log Routers. It happens that we have satellites
that contain the same mutations with Log Router tags, that have no other
peeking load, so we can prefer to use the satellite to peek rather than
the primary to distribute load across TLogs better.
Unfortunately, this revealed a latent bug in how tagged mutations in the
KnownCommittedVersion->RecoveryVersion gap were copied across
generations when the number of log router tags were decreased.
Satellite TLogs would be assigned log router tags using the
team-building based logic in getPushLocations(), whereas TLogs would
internally re-index tags according to tag.id%logRouterTags. This
mismatch would mean that we could have:
Log0 -2:0 ----- -2:0 Log 0
Log1 -2:1 \
>--- -2:1,-2:0 (-2:2 mod 2 becomes -2:0) Log 1
Log2 -2:2 /
And now we have data that's tagged as -2:0 on a TLog that's not the
preferred location for -2:0, and therefore a BestLocationOnly cursor
would miss the mutations.
This was never noticed before, as we never
used a satellite as a preferred location to peek from. Merge cursors
always peek from all locations, and thus a peek for -2:0 that needed
data from the satellites would have gone to both TLogs and merged the
results.
We now take this mod-based re-indexing into account when assigning which
TLogs need to recover which tags from the previous generation, to make
sure that tag.id%logRouterTags always results in the assigned TLog being
the preferred location.
Unfortunately, previously existing will potentially have existing
satellites with log router tags indexed incorrectly, so this transition
needs to be gated on a `log_version` transition. Old LogSets will have
an old LogVersion, and we won't prefer the sattelite for peeking. Log
Sets post-6.2 (opt-in) or post-6.3 (default) will be indexed correctly,
and therefore we can safely offload peeking onto the satellites.
The redundant team removed by teamRemover will not exist
in the global teams data structure. So we will not find
the redundant team from shard-to-team mapping in the system key.
Before this change, teamTracker marks such team as PRIORITY_TEAM_UNHEALTHY.
With this change, it marks it as PRIORITY_TEAM_REDUNDANT
A.J. Beamon