imlement issue #480 : UDP target host address is cached

[lxhoan]

this PR is a simpler solution for #480

[lxhoan]

Hi @fmachado , @majst01 , in case we have dns resolving cached at JVM level, still need this PR to fix issue #480

[codecov-io]

Codecov Report

Merging #502 into master will decrease coverage by 0.08%.
The diff coverage is 100%.

@@             Coverage Diff              @@
##             master     #502      +/-   ##
============================================
- Coverage     87.53%   87.45%   -0.09%     
+ Complexity      365      364       -1     
============================================
  Files            25       25              
  Lines          1500     1506       +6     
  Branches        167      167              
============================================
+ Hits           1313     1317       +4     
- Misses          120      123       +3     
+ Partials         67       66       -1

Impacted Files	Coverage Δ	Complexity Δ
src/main/java/org/influxdb/impl/InfluxDBImpl.java	89.97% <100%> (+0.16%)	78 <2> (ø)	⬇️
src/main/java/org/influxdb/InfluxDBException.java	87.27% <0%> (-3.64%)	11% <0%> (-1%)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 24c5542...9434330. Read the comment docs.

[lxhoan]

Hi @majst01 , @fmachado
What do you think ?
tks

[majst01]

Hi @lxhoan

I like this approach, but we should at least measure what performance impact this gives ?

[fmachado]

Except for my last comment, I'm fine with merging this PR.

[fmachado]

Is HttpUrl able to parse URIs that do not start with the string http? IMO we should treat this parameter as an URI and not an URL. The main reasons are:

• Users may finally stop using the scheme "http://" to connect via UDP. RFC 1738 for URL does not allow the scheme "udp://" but the RFC 3986 for URI does.
• If so, this line could be replaced with URI.create(...) and later you could use URI#getHost() to get only the hostname.

[lxhoan]

@fmachado , I have fixed as your comment, tks

[lxhoan]

Hi @majst01 ,

I like this approach, but we should at least measure what performance impact this gives ?

Performance impact should be inconsiderable since DNS resolving is cached at JVM level (as we discussed intensively before), so most of InetSocketAddress resolving is are just intra-process invocations