[
https://issues.apache.org/jira/browse/HAWQ-323?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15086862#comment-15086862
]
zharui commented on HAWQ-323:
-----------------------------
After set net.ipv4.tcp_timestamps=1 and net.ipv4.tcp_tw_recycle=1, the
Segmentation fault is disappear, but the problem about communication between
segment and master resource manager still exists. I ran "netstat -apn | grep
5437" and the results as follows
tcp 0 0 0.0.0.0:5437 0.0.0.0:*
LISTEN 6409/postgres
tcp 0 1 192.168.3.2:12877 192.168.3.2:5437
SYN_SENT 6480/postgres
tcp 0 1 192.168.3.2:5437 192.168.3.3:21157
LAST_ACK -
tcp 0 1 192.168.3.2:12881 192.168.3.2:5437
SYN_SENT 6480/postgres
tcp 0 1 192.168.3.2:12879 192.168.3.2:5437
SYN_SENT 6480/postgres
tcp 0 1 192.168.3.2:5437 192.168.3.3:21217
LAST_ACK -
I see that LAST_ACK is to 192.168.3.3 but not ack to 192.168.3.2. BTW, process
6480 is the segment resource menager on the same node with master.
> Cannot query when cluster include more than 1 segment
> -----------------------------------------------------
>
> Key: HAWQ-323
> URL: https://issues.apache.org/jira/browse/HAWQ-323
> Project: Apache HAWQ
> Issue Type: Bug
> Components: Core, Resource Manager
> Affects Versions: 2.0.0-beta-incubating
> Reporter: zharui
> Assignee: Lei Chang
>
> The version I use is 2.0.0-beta-RC2. I can query data normally when cluster
> just have 1 segment. Once the cluster have more then 1 segments online, I
> cannot finish any query and being informed that "ERROR: failed to acquire
> resource from resource manager, 7 of 8 segments are unavailable
> (pquery.c:788)".
> I have read the segment logs and the source code about resource manager. I
> guess this issue is because of the communication failure between segment
> instance and resource manager server. I can find the logs of the segment
> connect to resource manager successfully such as "AsyncComm framework
> receives message 518 from FD5" and "Resource enforcer increases memory quota
> to: total memory quota=65536 MB, delta memory quota = 65536 MB", but the
> other online segments have no these log.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
