|Pivotal Greenplum (GPDB)||4.3.x, 4.2.x|
|Pivotal Greenplum Command Center (GPCC)||1.x|
|EMC Data Computing Appliance||184.108.40.206|
After DCA upgrade to v2040, the Pivotal Greenplum Command Center shows an empty page for storage monitoring.
The Command Center uses gpsmon to track the real hostname and the GPDB catalog table to check segment configuration. Once the real hostname changes, there is an inconsistency between these two sources. GPCC would run a query like the one listed below to check space usage :
max(sample_time) as sample_time,
count(distinct(hostname)) as hosts_current,
hosttype as hostname,
SUM(total_bytes) as total_bytes,
sum(bytes_used) as bytes_used,
sum(bytes_available) as bytes_available
(select sample_time, hostname, filesystem, hosttype, oid,
round((total_bytes::real/1073741824)::numeric, 2) as total_bytes, round((bytes_used::real/1073741824)::numeric, 2) as bytes_used,
round((bytes_available::real/1073741824)::numeric, 2) as bytes_available
(select ctime as sample_time, * from diskspace_history
where ctime in (select distinct max(ctime) from diskspace_history )) a,
(select distinct(hostname) as customerhostname,
case content when -1 then 'master' else 'sdw' end as hosttype from gp_segment_configuration group by hostname, content) gseg
mount = filesystem
and hostname = customerhostname
and hosttype in ('master', 'sdw')) as c
group by hosttype, oid;
This query will read the hostname from three tables:
- Manually change the hostname of each server to the same hostname in the gp_segment_configuration table (update /etc/sysconfig/network file with root account).
- Then truncate diskspace_history table in gpperfmon db.
- Finally, restart the database to ensure that gpsmon reads the correct hostname value.