Bug 889231

Description of problem:
Some operations on spacewalk cause a query in the database to hang in this state:
postgres: spaceuser spaceschema 192.168.14.10(57927) idle in transaction

It will stay there forever until manually killed.
It happens with at least one specific centos5 server when running rhn_check

Bellow are the queries executed on the DB:

Q...:BEGIN; SET TRANSACTION ISOLATION LEVEL READ COMMITTED.Q...
select 1.Q.../select label, value from rhnTemplateString.Q...`
        select s.id , s.secret from rhnServer s 
        where s.id = 1000010105 
        .Q...V
        select s.id  from rhnServer s 
        where s.id = 1000010105 
        .Q...2select * from rhnServer where id = 1000010105.Q...F
        select label from rhnServerArch where id = 1023
        .Q....
    update rhnServerInfo
    set checkin = current_timestamp, checkin_counter = checkin_counter + 1
    where server_id = 1000010105
    .Q....COMMIT.Q...:BEGIN; SET TRANSACTION ISOLATION LEVEL READ COMMITTED.Q....
        select TO_CHAR(expires, 'YYYY-MM-DD HH24:MI:SS') expires
          from rhnSatelliteCert
         where label = 'rhn-satellite-cert'
         order by version desc nulls last
        .Q...Wselect server_id, label from rhnServerEntitlementView where server_id = 1000010105.Q....
            select at.label action_type,
                   at.trigger_snapshot,
                   at.name
              from rhnServerAction sa,
                   rhnAction a,
                   rhnActionType at
             where sa.server_id = 1000010105 
               and sa.action_id = 6731 
               and sa.status = 1
               and a.id = 6731
               and a.action_type = at.id
        .Q....
    update rhnServerAction
        set status = 2,
            result_code = 0,
            result_msg  = E'Files successfully diffed',
            completion_time = current_timestamp
    where action_id = 6731
      and server_id = 1000010105
    .Q....
    update rhnActionConfigRevision
       set failure_id = NULL
     where server_id = 1000010105
       and action_id = 6731
.Q...i
    select acr.id, cfn.path
      from rhnConfigFileName cfn,
           rhnConfigFile cf,
           rhnConfigRevision cr,
           rhnActionConfigRevision acr
     where acr.server_id = 1000010105
       and acr.action_id = 6731
       and acr.config_revision_id = cr.id
       and cr.config_file_id = cf.id
       and cf.config_file_name_id = cfn.id
.Q...@select * from rhnConfigFileFailure where label = E'missing'.Q...Y
    update rhnActionConfigRevision
       set failure_id = 1
     where id = 98849
.Q...Y
    update rhnActionConfigRevision
       set failure_id = 1
     where id = 98673
.Q...e
    select acr.id
      from rhnActionConfigRevision acr
     where acr.server_id = 1000010105
This was obtained with wireshark, so there are some garbage characters in there.
After the last select statement there is no further communication and a commit is not issued.
That select returns 49000 records.

Version-Release number of selected component (if applicable):
1.8 with Psql backend

How reproducible:
Always for a specific server when running rhn_check.
Occasionally on other servers that run centos5 when running rhn_check

Curiously, never happens with centos6 servers.

Steps to Reproduce:
1.
2.
3.
  
Actual results:


Expected results:


Additional info: