You could also prove this with prometheus 2.13.0, by setting environment 
variable GODEBUG=tls13=0 and see if it starts to work.

(That workaround was removed in later versions of go, so it's not usable in 
production - but at least it will demonstrate whether this is the problem 
or not)

On Monday, 12 September 2022 at 10:06:54 UTC+1 Brian Candler wrote:

> One problem that affected me was that go version 1.15+ removed the ability 
> to use a certificate which has only "commonName" 
> <https://golang.org/doc/go1.15#commonname>.  It now requires 
> "subjectAltName" to be present. That, I believe, would have affected 
> prometheus 2.21.0 
> <https://github.com/prometheus/prometheus/releases/tag/v2.21.0> onwards, 
> so it doesn't explain your change at 2.13.0.
>
> Prometheus 2.12.0 was built with go1.12.8, and 2.13.0 was built 
> with go1.13.1, so it could be a similar issue.
>
> Now, looking at the go1.13 release notes 
> <https://go.dev/doc/go1.13#tls_1_3>, I see that TLS 1.3 was enabled by 
> default then - there's more information about go's TLS 1.3 support in the 
> go1.12 
> release notes <https://go.dev/doc/go1.12#tls_1_3> when it was made 
> available but opt-in only.
>
> Perhaps that's the issue - your target server has a broken TLS 1.3 
> implementation?  If so, forcing TLS 1.2 might be a workaround.  
> Unfortunately, it looks like prometheus' tls_config 
> <https://prometheus.io/docs/prometheus/latest/configuration/configuration/#tls_config>
>  in 
> scrape jobs allows setting a minimum TLS version, but not a  maximum :-(
>
> That's weird, because when configuring prometheus' own https endpoint 
> <https://prometheus.io/docs/prometheus/latest/configuration/https/>, you 
> can select both min and max versions.
>
> Maybe disabling TLS 1.3 in the exporter itself is an option?
>  
> On Monday, 12 September 2022 at 09:23:35 UTC+1 [email protected] wrote:
>
>> Hi Brian,
>> thanks for these tips. I already tried those before my post, but they 
>> didn't help.
>> In the meantime I did some more investigation. The servers creating the 
>> metrics are Payara Micro 5.2021.10
>> The Prometheus (actually Thanos) is v2.38.0. Strangly it works with 
>> 2.12.0 (old on prem installation that got replaced), and is broken with 
>> every version starting at 2.13.0.
>> Should be quite obvious to figure out, but I don't see any changes in 
>> 2.13.0 that I would pinpoint this too.
>>
>> I asked the guys running the servers if they could look into their 
>> config. e.g. disabling h2 didn't help either.
>>
>> Very strange issue :(
>>
>> Cu
>>  Markus
>>
>> On Friday, 9 September 2022 at 15:10:46 UTC+2 Brian Candler wrote:
>>
>>> Note: the ability to disable http2 via scrape config was only added in 
>>> v2.35.0 <https://github.com/prometheus/prometheus/releases/tag/v2.35.0>.
>>>
>>> For an older version, you could try environment variables
>>> DISABLE_HTTP2=1
>>> and/or
>>> GODEBUG=http2client=0
>>>
>>> On Friday, 9 September 2022 at 13:18:42 UTC+1 Brian Candler wrote:
>>>
>>>> What prometheus version are you using?  I'd suggest v2.37.0 (2.37 is an 
>>>> LTS release branch)
>>>>
>>>> "promtool debug metrics" doesn't seem to have many options, but you 
>>>> could try in your prometheus scrape config:
>>>>
>>>> enable_http2: false
>>>>
>>>> and/or
>>>>
>>>> tls_config:
>>>>   insecure_skip_verify: true
>>>>
>>>> to try and narrow down the problem.
>>>>
>>>> On Friday, 9 September 2022 at 10:19:40 UTC+1 [email protected] wrote:
>>>>
>>>>> promtool debug metrics https://myserver:8181
>>>>>
>>>>> collecting: https://myserver:8181/metrics
>>>>>
>>>>> 2022/09/09 11:15:05 http2: Transport failed to get client conn for 
>>>>> myserver:8181: http2: no cached connection was available
>>>>>
>>>>> error completing debug command: error executing HTTP request: Get "
>>>>> https://myserver:8181/metrics": EOF
>>>>>
>>>>>
>>>>> If I open the url in a browser or with curl I get proper response :( 
>>>>> Any hint what I might be missing? testssl.sh works fine too
>>>>>
>>>>>
>>>>>
>>>>> On Friday, 9 September 2022 at 10:37:13 UTC+2 Markus Glück wrote:
>>>>>
>>>>>> Hi guys,
>>>>>> I currently trying to debug a similar issue. I am also getting EOF 
>>>>>> only as error in debug level from scrape manager. I had the same issue 
>>>>>> locally with curl and LibreSSL on macOS. Updating the curl version to 
>>>>>> use 
>>>>>> openSSL fixed it. So my assumption it's something related to SSL/TLS. My 
>>>>>> working curl verbose output gives me this in regards to TLS:
>>>>>>
>>>>>> * ALPN: offers h2
>>>>>>
>>>>>> * ALPN: offers http/1.1
>>>>>>
>>>>>> } [5 bytes data]
>>>>>>
>>>>>> * TLSv1.3 (OUT), TLS handshake, Client hello (1):
>>>>>>
>>>>>> } [512 bytes data]
>>>>>>
>>>>>> * TLSv1.3 (IN), TLS handshake, Server hello (2):
>>>>>>
>>>>>> { [193 bytes data]
>>>>>>
>>>>>> * TLSv1.3 (OUT), TLS change cipher, Change cipher spec (1):
>>>>>>
>>>>>> } [1 bytes data]
>>>>>>
>>>>>> * TLSv1.3 (OUT), TLS handshake, Client hello (1):
>>>>>>
>>>>>> } [512 bytes data]
>>>>>>
>>>>>> * TLSv1.3 (IN), TLS handshake, Server hello (2):
>>>>>>
>>>>>> { [155 bytes data]
>>>>>>
>>>>>> * TLSv1.3 (IN), TLS handshake, Encrypted Extensions (8):
>>>>>>
>>>>>> { [51 bytes data]
>>>>>>
>>>>>> * TLSv1.3 (IN), TLS handshake, Certificate (11):
>>>>>>
>>>>>> { [4943 bytes data]
>>>>>>
>>>>>> * TLSv1.3 (IN), TLS handshake, CERT verify (15):
>>>>>>
>>>>>> { [520 bytes data]
>>>>>>
>>>>>> * TLSv1.3 (IN), TLS handshake, Finished (20):
>>>>>>
>>>>>> { [52 bytes data]
>>>>>>
>>>>>> * TLSv1.3 (OUT), TLS handshake, Finished (20):
>>>>>>
>>>>>> } [52 bytes data]
>>>>>>
>>>>>> * SSL connection using TLSv1.3 / TLS_AES_256_GCM_SHA384
>>>>>>
>>>>>> * ALPN: server accepted h2
>>>>>>
>>>>>>
>>>>>> but HOW do I get more info from Prometheus?
>>>>>>
>>>>>> Thx,
>>>>>>
>>>>>>   Markus
>>>>>>
>>>>>> On Thursday, 14 July 2022 at 16:25:36 UTC+2 Brian Candler wrote:
>>>>>>
>>>>>>> The straightforward way to debug this by doing a scrape by hand: 
>>>>>>> it's just a HTTP request.
>>>>>>>
>>>>>>> curl -g 'https://blah.local:9126/metrics'
>>>>>>>
>>>>>>> Add flag '-v' for more debugging if required (e.g. response headers 
>>>>>>> may give you an extra clue).  Once you're able to scrape the exporter 
>>>>>>> with 
>>>>>>> curl, then prometheus should be able to talk to it too.
>>>>>>>
>>>>>>> Given that it's https, if you get a certificate error then you can 
>>>>>>> add flag '-k' to skip certificate verification.  If that turns out to 
>>>>>>> be 
>>>>>>> the problem, then there are extra flags you can pass to curl, e.g. to 
>>>>>>> pass 
>>>>>>> the CA root certificate (if the target's cert was signed by a CA that's 
>>>>>>> not 
>>>>>>> in the system trust store).  Once you've got all that working, you can 
>>>>>>> make 
>>>>>>> the corresponding changes to prometheus' tls_config.
>>>>>>>
>>>>>>> On Thursday, 14 July 2022 at 14:46:24 UTC+1 [email protected] 
>>>>>>> wrote:
>>>>>>>
>>>>>>>> Hello Team, 
>>>>>>>>
>>>>>>>> I have promethus installed on kubernet and add scraping targets 
>>>>>>>> from cmdb using http_sd_config. 
>>>>>>>>
>>>>>>>> I am getting EOF error while i add targets. And it’s show down. 
>>>>>>>>
>>>>>>>> Can someone please help me what is the issue. 
>>>>>>>>
>>>>>>>> For your reference attached screenshot of error. 
>>>>>>>>
>>>>>>>> Thanks and regards
>>>>>>>> Ritesh patel 
>>>>>>>>
>>>>>>>

-- 
You received this message because you are subscribed to the Google Groups 
"Prometheus Users" 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/prometheus-users/449b8ce5-b262-4a95-a564-5057bc46d60bn%40googlegroups.com.

Reply via email to