Re: How to 'clear' the full log dbspace?
Posted in 1997
In article <19970121045501.XAA26412@ladder01.news.aol.com>, vfraenkle@aol.com says... > >Im Artikel <5buu1g$16o@cssun.mathcs.emory.edu>, xiaodong@secom.co.cn >("alex") schreibt: > >>I'm using Informix Online 5.01 on the SCO 3.2.4. >> >>Yesterday I executed an big transaction process that exceeded my log >>volume, the log was full and everything halted, then I killed the process >>and use 'tbstat -l' to check my log info, finding that almost all log >space >>are 100% full, when I use 'tbtape -s' to backup(the tape device pointed >to >>/dev/null) the log, but the program waits forever... and online server >>could not be brought to online or quiescent mode, only the recovery mode. >> >>Under this situation, how can I init the logspace and bring the server to >>normal state? >> >> >> > >Hello alex, >in this situation, your only is to throw away your dbspaces, build new >ones and restore your database(s) from the tape. (This is, what Informix >support sayed to us). > >In some releases of INFORMIX 5 there is a undocumented Flag for tbinit. >Try tbinit -F. This will bring your online up, without restoring from the >logical logs. >WARNING: After that, your database may be inconsistent! > >After restoring your database, please set the following parameters to the >described values (in your tbconfig file): > > LTXHWM=60 > LTXEHWM=70 > >This prevents your database for the next time. > >Hope that helps, >Volker Fraenkle Hi, If you do not have Openline Maintainance agreement then yes, the only way is to reinitialize the whole thing and restore from archive (if one is available ;-) ) Informix themselves can add some logical space dynamically to allow your long transaction to rollback .That is if you have Online Agreement with them. Concerning setting upt LTXHWM and LTXEHWM mentioned above I should say this settings are bloody application specific, I usually recommend to set it to 45 and 47 respectively plus make sure you have got enough log space. On other hand I personally witnessed this condition with the thresholds set to 35 and 40 with logical logs set up nicely. Cheers, Ilya Bogorad DHL Systems Ltd