Vault 6.0, 6.1
What are the INI settings possible for indexerd?
The behavior of the indexer as a service can be customized by creating an initialization file.
Controls how many concurrent threads are created to process the requests made to
the Indexer as a Service. Additional threads may improve performance of searches
during heavy load times (for example when loading new job data).
The thread count should be at least one and no larger than the number of real CPU
cores less 2. So a quad core system should have one or two threads set. If the platform
is doing additional work besides handling Vault or additional Vault servers and/or
renders have been set up on the same platform as the indexer, then the thread count
should be reduced accordingly.
As each thread can consume the processing power of a processing core, ThreadCount
should be increased if your processing platform has the addition processing capacity
(unused processor core). Increasing the ThreadCount in a system where the
additional capacity does not exist (or is used by other processes) can actually result
in lower indexing throughput due to excess contention for processing capacity.
Controls how much memory is allocated to the index cache (in MB). A higher cache
will generally speed up index processing but will also consume more of the system’s
The default setting is 1024 MB and this is the maximum for 32 bit systems. The cache
can be reduced for systems that are short of processor memory but indexer
performance will be affected.
Controls how often the indexer will write out all the changed indexes in the index
cache to the index file (in minutes). The Indexer as a Service will write out changed
indexes to the index file when it is not servicing requests. A full flush (writing out all
changed indexes before servicing requests to add or change another index) will occur
at every FullFlush interval.
The default value is 1440 minutes.
FullFlush=0 means that a full flush will occur at the end of every Vault job load.
Note: A FullFlush value of 0 can have a large impact on indexing performance when
you are doing an index or reindex operation for a large of number of jobs in a single
The FlushPeriod sets the number of seconds an updated (or new) index entry in the
index cache needs to be unchanged for before it is flushed (written from the index
cache to the index file and its cache space made available for reuse).
The default value is 60.
A longer FlushPeriod will mean that it takes longer period for the index file to be
updated with the cache contents.
QueueSize sets the size of the indexer command queue. If the command is too small
and fills up, this can adversely affect the rate at which indexer events can be
processed. However, increasing the comamnd queuesize will consume more memory
resources within your system. A proper setting will require some experimentation to
adjust for your system memory availability and your usage pattern.
The default value is 200000.
The minimum value is 10000 and the maximum value allowed is 1200000.
An oscache value of 1 will cause the Windows version of the indexer to use unbuffered
writes to update the index file. This setting can be used in low memory environments
where the Windows O/S is caching the index file updates and consuming excessive
system memory due to this caching.
oscache has no effect in AIX/Solaris/Linux versions of the indexer.
Note: Setting oscache=1 may actually slow down the indexer operation if there is
sufficient memory for it to operate correctly without oscache=1 being set.
UPDATED: June 28, 2017