MySQL Bugs: #42559: More than 32 index/tup scan on one fragment can cause timeouts (or lcp-hang)

Bug #42559	More than 32 index/tup scan on one fragment can cause timeouts (or lcp-hang)
Submitted:	2 Feb 2009 21:06	Modified:	3 Feb 2009 12:34
Reporter:	Jonas Oreland	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Cluster: Cluster (NDB) storage engine	Severity:	S3 (Non-critical)
Version:		OS:	Any
Assigned to:	Jonas Oreland	CPU Architecture:	Any

Description:
If doing more than 32 index *or* tup-scan on one fragment
scans could be left hanging too long, this can
cause unnecessary timeouts, and/or worst case hanging LCP

this was introduced by fix for bug#42084

problem was that they share ScanNo - range (used for ScanTakeOver)

and when an index-scan finishes it will only start queued index-scans
(or when an tup-scan finishes it will only start queued tup-scans)

solution is to split the ScanNo range
(so that acc scan has one range, tux scan has one range and tup scan has one range)

How to repeat:
run testScan --l 1000 -n 488O T1
while running LCP on highest speed, note that it doesnt run continuously

Suggested fix:
split range

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/64939

2806 Jonas Oreland	2009-02-02
      ndb - bug#42559

Pushed into 5.1.31-ndb-6.4.3 (revid:jonas@mysql.com-20090202213027-xfip15kq57fu9klj) (version source revid:jonas@mysql.com-20090202213027-xfip15kq57fu9klj) (merge vers: 5.1.31-ndb-6.4.3) (pib:6)

Pushed into 5.1.31-ndb-6.2.17 (revid:jonas@mysql.com-20090202212134-uhk6g48lz95m4nc5) (version source revid:jonas@mysql.com-20090202212134-uhk6g48lz95m4nc5) (merge vers: 5.1.31-ndb-6.2.17) (pib:6)

Pushed into 5.1.31-ndb-6.3.22 (revid:jonas@mysql.com-20090202212343-opce431gg4wqivq4) (version source revid:jonas@mysql.com-20090202212343-opce431gg4wqivq4) (merge vers: 5.1.31-ndb-6.3.22) (pib:6)

Documented in the NDB-6.2.17, 6.3.22, and 6.4.2 changelogs as follows:

        When performing more than 32 index or tuple scans on a single 
        fragment, the scans could be left hanging. This caused unnecessary 
        timeouts, and in addition could possibly lead to a hang of an LCP.

        This bug was introduced by the fix for Bug #42084.

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/65148

2810 Jonas Oreland	2009-02-04
      ndb - test prg for bug#42559

Pushed into 5.1.31-ndb-6.2.17 (revid:jonas@mysql.com-20090204123522-5g0fd31u7mu83gej) (version source revid:jonas@mysql.com-20090204123522-5g0fd31u7mu83gej) (merge vers: 5.1.31-ndb-6.2.17) (pib:6)

Pushed into 5.1.31-ndb-6.3.22 (revid:jonas@mysql.com-20090204124039-4239bv0cmqq7kr0v) (version source revid:jonas@mysql.com-20090204124039-4239bv0cmqq7kr0v) (merge vers: 5.1.31-ndb-6.3.22) (pib:6)

Pushed into 5.1.31-ndb-6.4.3 (revid:jonas@mysql.com-20090204124234-btvep0nz9e2vk030) (version source revid:jonas@mysql.com-20090204124234-btvep0nz9e2vk030) (merge vers: 5.1.31-ndb-6.4.3) (pib:6)

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/65578

3248 Martin Skold	2009-02-09 [merge]
      Merge

Pushed into 6.0.10-alpha (revid:alik@sun.com-20090210194937-s7xshv5l3m1v7wi9) (version source revid:tomas.ulin@sun.com-20090204130805-wb1egwdby2y0cnti) (merge vers: 6.0.10-alpha) (pib:6)