Bug 1292827

The heavy lifting in ExtraPackages is being done by this mode-query:

===
<mode name="extra_packages_systems_count"
      class="com.redhat.rhn.frontend.dto.SystemOverview">
  <query params="userid, orgid">
select s.id as id,
       s.name as name,
       1 as selectable,
       count(sp.name_id) as extra_pkg_count
       from rhnServer s
  join rhnUserServerPerms usp on (usp.server_id = s.id and
                                  usp.user_id = :userid)
  join rhnServerPackage sp on (s.id = sp.server_id)
       left outer join (select sc.server_id,
                               cp.package_id,
                               p.name_id,
                               p.evr_id,
                               p.package_arch_id
                          from rhnPackage p,
                               rhnServerChannel sc,
                               rhnServerPackage sp2,
                               rhnChannelPackage cp,
                               rhnUserServerPerms usp2
                         where cp.package_id = p.id
                           and cp.channel_id = sc.channel_id
                           and sc.server_id = usp2.server_id
                           and usp2.user_id = :userid
                           and sc.server_id = sp2.server_id
                           and sp2.name_id = p.name_id
                           and sp2.evr_id = p.evr_id
                           and sp2.package_arch_id = p.package_arch_id
                       ) scp on (scp.server_id = sp.server_id and
                                 sp.name_id = scp.name_id and
                                 sp.evr_id = scp.evr_id and
                                 sp.package_arch_id = scp.package_arch_id)
 where scp.package_id is null and
       s.org_id = :orgid
 group by s.name, s.id
 order by name, id
  </query>
===

The explain plan for that query (on postgres in this example, 250 systems registered, running as sat-admin) looks like so:

===
 GroupAggregate  (cost=33950.01..33950.03 rows=1 width=28)
   ->  Sort  (cost=33950.01..33950.02 rows=1 width=28)
         Sort Key: s.name, s.id
         ->  Nested Loop  (cost=28954.68..33950.00 rows=1 width=28)
               Join Filter: (usp.server_id = s.id)
               ->  Nested Loop  (cost=28954.68..33949.72 rows=1 width=21)
                     Join Filter: (sp.server_id = usp.server_id)
                     ->  Index Scan using rhn_usperms_sid_idx on rhnuserserverperms usp  (cost=0.00..20.11 rows=219 width=8)
                           Filter: (user_id = 1::numeric)
                     ->  Materialize  (cost=28954.68..33926.32 rows=1 width=13)
                           ->  Hash Left Join  (cost=28954.68..33926.32 rows=1 width=13)
                                 Hash Cond: ((sp.name_id = p.name_id) AND (sp.evr_id = p.evr_id) AND (sp.package_arch_id = p.package_arch_id) A
ND (sp.server_id = sc.server_id))
                                 Filter: (cp.package_id IS NULL)
                                 ->  Seq Scan on rhnserverpackage sp  (cost=0.00..2745.25 rows=148425 width=23)
                                 ->  Hash  (cost=28954.32..28954.32 rows=18 width=29)
                                       ->  Nested Loop  (cost=27123.41..28954.32 rows=18 width=29)
                                             Join Filter: (sc.server_id = usp2.server_id)
                                             ->  Nested Loop  (cost=27123.41..28949.19 rows=18 width=37)
                                                   Join Filter: (sc.channel_id = cp.channel_id)
                                                   ->  Nested Loop  (cost=27123.41..28913.29 rows=51 width=42)
                                                         ->  Merge Join  (cost=27123.41..28897.31 rows=57 width=29)
                                                               Merge Cond: ((sp2.evr_id = p.evr_id) AND (sp2.name_id = p.name_id) AND (sp2.pack
age_arch_id = p.package_arch_id))
                                                               ->  Sort  (cost=18539.49..18910.56 rows=148425 width=23)
                                                                     Sort Key: sp2.evr_id, sp2.name_id, sp2.package_arch_id
                                                                     ->  Seq Scan on rhnserverpackage sp2  (cost=0.00..2745.25 rows=148425 widt
h=23)
                                                               ->  Sort  (cost=8377.70..8503.08 rows=50150 width=21)
                                                                     Sort Key: p.evr_id, p.name_id, p.package_arch_id
                                                                     ->  Seq Scan on rhnpackage p  (cost=0.00..4462.50 rows=50150 width=21)
                                                         ->  Index Only Scan using rhn_sc_sid_cid_uq on rhnserverchannel sc  (cost=0.00..0.27 r
ows=1 width=13)
                                                               Index Cond: (server_id = sp2.server_id)
                                                   ->  Index Scan using rhn_cp_pid_idx on rhnchannelpackage cp  (cost=0.00..0.62 rows=7 width=1
1)
                                                         Index Cond: (package_id = p.id)
                                             ->  Index Scan using rhn_usperms_sid_idx on rhnuserserverperms usp2  (cost=0.00..0.27 rows=1 width
=8)
                                                   Index Cond: (server_id = sp2.server_id)
                                                   Filter: (user_id = 1::numeric)
               ->  Index Scan using rhn_server_oid_id_idx on rhnserver s  (cost=0.00..0.27 rows=1 width=23)
                     Index Cond: ((org_id = 1::numeric) AND (id = sp.server_id))
(37 rows)
===

Note three Seq Scans:

->  Seq Scan on rhnserverpackage sp  (cost=0.00..2745.25 rows=148425 width=23)
->  Seq Scan on rhnserverpackage sp2  (cost=0.00..2745.25 rows=148425 widt
h=23)
->  Seq Scan on rhnpackage p  (cost=0.00..4462.50 rows=50150 width=21)


Investigation continues.

Fun - see

  https://bugzilla.redhat.com/show_bug.cgi?id=1099938

for the last time this came up.

Based on the discussion in '9938, this may be a case of "we can't do any better".

One possible approach would be to make 'extra packages' be calculated as part of update_needed_cache, since it's affected by the exact same transitions. The result could be kept in a column in rhnServerInfo, and the page would just display what's there.

Yes, this would work. I'd however vote for a separate table and don't use rhnServerInfo table for this purpose.

We have re-reviewed this bug, as part of an ongoing effort to improve Satellite/Proxy feature and bug updates, review and backlog.

This is a low priority bug and has no currently open customer cases. While this bug may still valid, we do not see it being implemented prior to the EOL of the Satellite 5.x product. As such, this is being CLOSED DEFERRED. 

Closing now to help set customer expectations as early as possible. You are welcome to re-open this bug if needed.