Hi Yuri,
Ok! In this case, the documentation should be more clear :/ So on both the
device and server-side, I have to reach somehow the underlying TCP
connection and set the timeouts correspondingly. I assume that on the
device I have to use a custom dialer (grpc.WithContextDialer); and on the
service, I have to use a custom listener (net.ListenConfig{KeepAlive: 5 * 60
* time.Second}). Don't I?
Regards,
Lajos
[email protected] a következőt írta (2022. március 28., hétfő, 21:00:35
UTC+2):
> Hi Lajos,
>
> The gRPC keep-alive configuration is only for the HTTP/2 keep-alive, not
> for the TCP keep-alive.
>
> On Mon, Mar 28, 2022 at 4:17 AM Lajos Rancz <[email protected]> wrote:
>
>> Hi!
>>
>> I've run strace on the deivce, and I got:
>> setsockopt(3, SOL_TCP, TCP_KEEPINTVL, [15], 4) = 0
>>
>> So the device process sets the TCP keep alive to 15 sec.
>>
>> Lajos Rancz a következőt írta (2022. március 28., hétfő, 10:40:33 UTC+2):
>>
>>> Hi all,
>>>
>>> I'm developing an IoT application, both the server (runs on a Digital
>>> Ocean droplet) and the client device application (runs on an embedded
>>> Linux) are written in Go. The internet connection between the devices and
>>> the server is provided by LTE Cat M1.
>>> The application requires a live connection between the device and the
>>> server (because of real-time events), so I use server-side streaming:
>>> message DeviceWatchRequest {
>>> int32 token = 1;
>>> }
>>>
>>> message DeviceWatchResponse {
>>> int64 id = 1;
>>> uint32 type = 2;
>>> bytes data = 3;
>>> }
>>>
>>> service DeviceService {
>>> ...
>>> rpc Watch(DeviceWatchRequest) returns (stream DeviceWatchResponse);
>>> }
>>>
>>> The whole system works well, BUT the data usage is unacceptably huge, I
>>> send only a few bytes (DeviceWatchResponse.data) hourly. So, I've
>>> dumped the TCP traffic with the tcpdump utility.
>>> What I see is that the vast majority of data usage is due to TCP
>>> keepalive:
>>>
>>> [image: wireshark.png]
>>>
>>> So, a keep-alive operation consists of 4 packets, all four are 68 bytes
>>> long; eg. one keep-alive operation is 272 bytes long. There are 86400
>>> seconds in a day, so 86400/15 = 5760 keep-alive operations are done in a
>>> day; it means 5760*272 = 1 566 720 bytes ~ 1.5 MB/day!
>>> My whole monthly data usage could be that much, not my daily. As
>>> https://www.emnify.com/pricing: 1 MB/month is around 1
>>> EUR/month/device, 50 MB is 2.65 EUR/month/device.
>>> So, I've tried to set the keep-alive options on client side:
>>> conn, err := grpc.Dial(*addr, grpc.WithInsecure(),
>>> grpc.WithKeepaliveParams(keepalive.ClientParameters{Time: 60 * time.Second,
>>> Timeout: 5 * time.Second, PermitWithoutStream: true}))
>>> But after 3 minutes the connection is dropped. The keep-alive
>>> documentation (https://github.com/grpc/grpc/blob/master/doc/keepalive.md
>>> and https://pkg.go.dev/google.golang.org/grpc/keepalive) is
>>> contradictory.
>>>
>>> What can I do?
>>>
>>> Regards,
>>> Lajos
>>>
>>> --
>> You received this message because you are subscribed to the Google Groups
>> "grpc.io" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to [email protected].
>> To view this discussion on the web visit
>> https://groups.google.com/d/msgid/grpc-io/27f8b79d-3f62-4916-881c-0c7f50d4e0cfn%40googlegroups.com
>>
>> <https://groups.google.com/d/msgid/grpc-io/27f8b79d-3f62-4916-881c-0c7f50d4e0cfn%40googlegroups.com?utm_medium=email&utm_source=footer>
>> .
>>
>
--
You received this message because you are subscribed to the Google Groups
"grpc.io" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/grpc-io/55058246-01ca-4fb3-9028-9a630af35fafn%40googlegroups.com.
