After the upgrade of the test instance to the last version 4.8.91. We noticed a quite longer startup time of the MGM with respect to the previous version, We were wondering if this is normal or might we have an issue in the configuration. It is not a very long time and is quite reasonable for a restart, i.e. around 20s, but previously this was more around 4-5 s in the test instance, which has few files and few FSTs (3).
By startup time I mean the time after which the eos server responds to the commands (for instance eos ns or eos ls). This is not due to the namespace booting, which is displayed in eos ns as less than 2 s, but seems to be linked to the preparation of the different threads.
Do you think this might be normal, or might there be something that blocks it temporarily ? We wouldn’t like that with the upgrade of the much bigger production instance this time became too long.
Yet another quick question :
Is the variable EOS_USE_NEW_INODES still necessary when working with high file id numbers (we met this issue few years ago) or is it now the default ?