Re: Informix Mirroring

eherber wrote: > On 3 Feb., 19:04, Dave <in4mix...@gmail.com> wrote: >> Hello, >> >> I was reading up on Informix Disk/Chunk Mirroring and was wondering if >> anyone has any experience using Informix mirroring versus O/S or >> Hardware mirroring relating to performance? >> >> Also, is this a typo in the Informix documentation for mirroring? >> >> See the sentences, "On the other hand, when an mirrored chunk goes >> down, the database server cannot access the data stored on that chunk. >> If the chunk contains logical-log files, the physical log, or the root >> dbspace, the database server goes offline immediately." >> >> Can that possibly be ? >> >> Here is the link to the source Doc:http://publib.boulder.ibm.com/infocenter/idshelp/v115/topic/com.ibm.a... >> >> Full Text: >> Consequences of Not Mirroring >> If you do not mirror your dbspaces, the frequency with which you have >> to restore from a storage-space backup after media failure increases. >> >> When a mirror chunk suffers media failure, the database server reads >> exclusively from the chunk that is still online until you bring the >> down chunk back online. On the other hand, when an mirrored chunk goes >> down, the database server cannot access the data stored on that chunk. >> If the chunk contains logical-log files, the physical log, or the root >> dbspace, the database server goes offline immediately. If the chunk >> does not contain logical-log files, the physical log, or the root >> dbspace, the database server can continue to operate, but threads >> cannot read from or write to the down chunk. If an unmirrored chunk >> goes down, you must restore it by recovering the dbspace from a >> backup. > > My experience with Informix mirroring is consistently positive. > It is a very cost efficient and reliable solution. > > Regarding the documentation issue: > > The database server goes only down if the original dbspace/chunk > already > failed and IDS is already using the mirror chunk. If that one fails > as well and is considered a critical dbspace/chunk, IDS will abort. > I think that is what the documentation tries to say. I also think the sentence is badly written, but the idea should be what Eric wrote. A quick search did not find any doc-bug for this. At least it should be written in a way that did not originate this discussion... Regards.