Domino server crashes during full text index search or when queries return large result set
Your Domino server crashes during full-text searching or when queries are performed on search terms that have numerous result sets.
Fatal thread: nServer
[ 1] 0x7c342f57 MSVCR71.memcpy+156 (c107692,91c6813d,16,358cf9b0)
@[ 2] 0x11b2672f gtr40nts.gtr_InitPctlStemming+303 (18,c1076c4,36fa5574,c100002)
@[ 3] 0x11b29524 gtr40nts.gtr_GetOneWordExactMixInit+2612 (3589039c,358af6a8,36fa5574,36fa91d4)
@[ 4] 0x11b33570 gtr40nts.gtr_RtrvWords+880 (36fa5574,36fa91d4,0,0)
@[ 5] 0x11b33d47 gtr40nts.GTR_Search+711 (4e000002,32928b89,1,3589039c)
@[ 6] 0x11b55ed8 gtr40nts.GTRsearchIndex+2200 (32928b65,c10e1f0,c10e2f0,c10e3f0)
@[ 7] 0x11af2131 nftgtr40.CSearchGTR::ExecQuery+401 (c10f4d3,c10faf4,c10f48c,11af28ce)
@[ 8] 0x11af2336 nftgtr40.CSearchExec::Start+102 (53dcbb4,37423fa8,32a463b8,1fba515c)
@[ 9] 0x11af10ed nftgtr40.CSearch::run+509 (400219,37423fa8,1550000,61697428)
@ 0x11af17ca nftgtr40.FTGSearch@24+170 (32a463b8,37423fa8,0,53dcbb4,0,0)
@ 0x601e0ab5 nnotes.FTSearchExt@60+741 (f9,35577af2,53d02fc,53dcbb4,0,0,0,0,0,0,0,0,0,0,0)
@ 0x1005a1d2 nserverl.ServerFTSearch@8+1090 (324b7294,65480007)
@ 0x1001f889 nserverl.DbServer@8+2361 (9d000156,605c0016)
@ 0x10033ffb nserverl.WorkThreadTask@8+1611 (1520328,0)
@ 0x10001a2e nserverl.Scheduler@4+750 (0)
Resolving the problem
This issue was reported to Quality Engineering as SPR# DVDI7ED9EF and has been fixed in Domino releases 8.5.2.
NOTE: In some cases, this fix does not address the issue. If this is the case, as a temporary workaround, identify the problem database and delete the index.
As a preventative measure, enable the feature described in SPR MTMY5J2JX8 by setting the notes.ini "FTG_ENABLE_GTR_KEY_CHECK=1". This will identify corrupted data contained in any of the GTR KEY files.
In most cases, rebuilding the full text index resolves the issue. As most corrupted indexes are associated with very large databases, it is best to target the problem database for rebuild. From server console, run the following command: load updall -x <name of database>.nsf
As a workaround, some administrators deleted the full-text index of the affected mail files, and then did not recreate the index until the server was upgraded to a fixed release.
Requested steps to collect critical data during workaround:
1) Determine, if possible, what exact query was being performed at the time of the failure. Debug to capture such information would be "DEBUG_FTV_SEARCH=1".
2) Save off the full text index associated with the database where the query was performed by moving it to another location. As size has been a factor in previous occurrences, create an archive of the full text index directory and all its' contents.