Mailing-List: contact reviews-help@impala.incubator.apache.org; run by ezmlm
Precedence: bulk
Message-Id: <201707102305.v6AN5nkm026693@ip-10-146-233-104.ec2.internal>
Date: Mon, 10 Jul 2017 23:05:49 +0000
From: "Henry Robinson (Code Review)" <gerrit@cloudera.org>
To: John Russell <jrussell@cloudera.com>, impala-cr@cloudera.com,
        reviews@impala.incubator.apache.org
CC: Lars Volker <lv@cloudera.com>
Reply-To: henry@cloudera.com
Subject: =?UTF-8?Q?=5BImpala-ASF-CR=5D_IMPALA-4978_/_IMPALA-5631=3A_=5BDOCS=5D_Add_FQDN_known_issue=0A?=
In-Reply-To: <gerrit.1499710242000.Ib039d0102878f1c05470371f581cb258287b9bc0@gerrit.cloudera.org>
References: <gerrit.1499710242000.Ib039d0102878f1c05470371f581cb258287b9bc0@gerrit.cloudera.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Content-Disposition: inline
User-Agent: Gerrit/2.12.7
archived-at: Mon, 10 Jul 2017 23:24:07 -0000

Henry Robinson has posted comments on this change.

Change subject: IMPALA-4978 / IMPALA-5631: [DOCS] Add FQDN known issue
......................................................................


Patch Set 2:

(1 comment)

After some discussion, I think two issues are being conflated here.

1. If a complete KRB principal is not supplied via --principal, Impala tries to guess at the hostname part using gethostname(). This might not return the FQDN which is needed for Kerberos. The linked bug (IMPALA-4978) is really talking about this issue.

2. However, the more important issue is that IMPALA-5253 made it critical that for TLS-enabled deployments, every service thinks its hostname is the FQDN. That hostname is controlled by --hostname. If that is not set, Impala falls back to gethostname() - and this is where IMPALA-4978 comes in, because it should also use the getaddrinfo() call first. The workaround for this issue is to set --hostname correctly to the FQDN (or ensure that 'hostname' returns the FQDN by configuring name resolution correctly). We found this at Cloudera because in some situations our cluster management software would set --hostname to the shortname.

http://gerrit.cloudera.org:8080/#/c/7388/2/docs/topics/impala_known_issues.xml
File docs/topics/impala_known_issues.xml:

PS2, Line 156: This problem might occur immediately after an upgrade, due to changes in cluster
             :           management software that supplies the <codeph>--hostname</codeph> flag
             :           automatically to the Impala-related daemons.
It's a bit hard to understand *what* kind of changes would cause this, and without that knowledge this is a bit vague. In my opinion it makes sense to either link directly to the CM known issue (I think it's perfectly ok to point to third-party software that breaks Impala during upgrades), or to remove this entirely.


-- 
To view, visit http://gerrit.cloudera.org:8080/7388
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: Ib039d0102878f1c05470371f581cb258287b9bc0
Gerrit-PatchSet: 2
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: John Russell <jrussell@cloudera.com>
Gerrit-Reviewer: Henry Robinson <henry@cloudera.com>
Gerrit-Reviewer: John Russell <jrussell@cloudera.com>
Gerrit-Reviewer: Lars Volker <lv@cloudera.com>
Gerrit-HasComments: Yes