Query servers continue to run while the write server issues DDL operations. The table version log (TLV log) stores information about DDL operations and communicates information about new table versions to the query servers. The write server preserves older table versions for as long as needed. In single-node mode, no version logging takes place. Query servers cannot run and must be synchronized to restart.
The -iqmpx_sn 1 server startup option starts the server in single-node mode.
WARNING! You must specify the override startup switch (-iqmpx_ov 1) and start in single-node mode (-iqmpx_sn 1) when starting a multiplex write server after any failure. Never use multiplex mode (the default) for recovery.
If a server runs out of space during a checkpoint operation, try restarting in single-node mode. For example, if a server ran out of space and was killed, it could precipitate a rare situation where normal recovery failed. Suppose that many versions were created at the server and they were all retained because the query server used an older version or there was a dbremote delay. If the write server had too many recovered transactions to checkpoint, it could run out of space during the recovery checkpoint. Subsequently, users could not connect or add dbspaces. Starting in single-node mode deallocates retained versions and allows checkpoints to succeed.
