dns resolving failure

[Asher Tarnopolski]

hi folks, 

i've discovered lately that we are randomly getting "Search domain query failed. Exceeded max queries per resolve 4" 

coming all the way from netty for one of the URLs we are calling from http client. 

the interesting thing is that these failures are not consistent, and we catch this behavior like every 20th or 30th call.

i found this: https://github.com/eclipse/vert.x/issues/1479 but it's marked as solved for a while (we are on 3.5.1).

i understand that -Dvertx.disableDnsResolver=true should solve this, but is there a chance there is still a bug worth opening an issue?

cheers

 

[Thomas SEGISMONT]

There might be another issue... Is there any chance you could capture the dns request from Netty's resolver and then the same with stock resolver?

--
You received this message because you are subscribed to the Google Groups "vert.x" group.
To unsubscribe from this group and stop receiving emails from it, send an email to vertx+unsubscribe@googlegroups.com.
Visit this group at https://groups.google.com/group/vertx.
To view this discussion on the web, visit https://groups.google.com/d/msgid/vertx/48a68e29-a8b0-44c1-b729-0a0d8ed9d3f3%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Julien Viet]

have you configured search domains on your client ?

have you tried 3.5.2.CR1 or CR2 which updates the Netty version and has fixes in the DNS resolver ?

--
You received this message because you are subscribed to the Google Groups "vert.x" group.

To unsubscribe from this group and stop receiving emails from it, send an email to vertx+un...@googlegroups.com.

[Asher Tarnopolski]

Thomas, I'll check if we can do it. Can you elaborate about the differences between netty's and stock resolvers? 

Julien, haven't tried 3.5.2 yet.

[Thomas SEGISMONT]

I mean the difference between the resolver based on Netty's code and the resolver from the JVM (which you activate with vertx.disableDnsResolver)

2018-05-28 17:01 GMT+02:00 Asher Tarnopolski <ata...@gmail.com>:

Thomas, I'll check if we can do it. Can you elaborate about the differences between netty's and stock resolvers? 

Julien, haven't tried 3.5.2 yet.

--

You received this message because you are subscribed to the Google Groups "vert.x" group.

To unsubscribe from this group and stop receiving emails from it, send an email to vertx+unsubscribe@googlegroups.com.

Visit this group at https://groups.google.com/group/vertx.

To view this discussion on the web, visit https://groups.google.com/d/msgid/vertx/c0d00acc-7d80-45cd-b1a2-1af9dc6f6653%40googlegroups.com.

[Asher Tarnopolski]

yeh, i got that. i was wondering why they would work differently.

[Thomas SEGISMONT]

Ha, sorry no. I'm not knowledgeable with any of the resolver's internals. Julien should know better.

2018-05-28 18:36 GMT+02:00 Asher Tarnopolski <ata...@gmail.com>:

yeh, i got that. i was wondering why they would work differently.

--
You received this message because you are subscribed to the Google Groups "vert.x" group.
To unsubscribe from this group and stop receiving emails from it, send an email to vertx+unsubscribe@googlegroups.com.
Visit this group at https://groups.google.com/group/vertx.

To view this discussion on the web, visit https://groups.google.com/d/msgid/vertx/a58e45cd-39ed-462f-8347-b7a505874b1c%40googlegroups.com.

[Julien Viet]

they work differently but now the Netty's resolver is getting similar behavior than the JVM one (https://github.com/netty/netty/pull/6937)

Julien

> On 28 May 2018, at 18:36, Asher Tarnopolski <ata...@gmail.com> wrote:
>
> yeh, i got that. i was wondering why they would work differently.
>

> --
> You received this message because you are subscribed to the Google Groups "vert.x" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to vertx+un...@googlegroups.com.

> Visit this group at https://groups.google.com/group/vertx.

> To view this discussion on the web, visit https://groups.google.com/d/msgid/vertx/a58e45cd-39ed-462f-8347-b7a505874b1c%40googlegroups.com.

[Martin Traverse]

Hello,

So, I've come across a similar issue that still exists with Vertx 3.9 / Netty 4.1.48. It seems that in some circumstances, the JNDI mechanism can give out details of stale DNS servers, i.e. servers belonging to networks the machine is no longer connected to. If these appear at the top of the list and are unreachable, DNS resolution fails. I think a good place to fix this would be in the Netty DNS bootstrap mechanism, right after it picks up the default set of servers from JNDI, by applying some sort of prioritise and filter method.

I have raised a ticket with Netty here:

https://github.com/netty/netty/issues/10264

In the mean time it is possible to work around the issue by getting the default set of servers from Netty, checking whether they are reachable and passing only the reachable ones to Vertx as an option when it starts up. I put some details on StackOverflow here:

https://stackoverflow.com/questions/42110790/vertx-httpclient-getnow-not-working/61696414#61696414

I'm not sure whether it's worth doing anything about this at the Vertx level, possibly wait and see if Netty agree it should be fixed there first and if not then put something in Vertx. My feeling is that at least by the time it gets to client code it should be working without needing to do anything extra.

Hope this is helpful!

Martin.

On Monday, 28 May 2018 17:59:57 UTC+1, Julien Viet wrote:

they work differently but now the Netty's resolver is getting similar behavior than the JVM one (https://github.com/netty/netty/pull/6937)

Julien


> On 28 May 2018, at 18:36, Asher Tarnopolski <ata...@gmail.com> wrote:
>
> yeh, i got that. i was wondering why they would work differently.
>
> --
> You received this message because you are subscribed to the Google Groups "vert.x" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to ve...@googlegroups.com.