Hi,
I am testing a configuration that should sort files into different
archive ids based on location and I want to use a different parameters
for the copytool based on file size. But I think I ran into a bug in
that configuration.
My understanding is that the parameters should be done based on the
following order:
1) As default parameters for a policy
2) In policy triggers
3) In policy rules
4) In fileclass definitions
So these are the different levels in my example:
1) The parameters sections has a default and defines the command:
lhsm_archive_parameters {
nb_threads = 8;
## archive 1000 files max at once
# max_action_count = 1000;
# max_action_volume = 1TB;
# suspend policy run if action error rate > 50% (after 100 errors)
suspend_error_pct = 50%;
suspend_error_min= 100;
# overrides policy default action
action = cmd("lfs hsm_archive --archive {archive_id} --data {data}
/gdatat1/.lustre/fid/{fid}");
# default action parameters
action_params {
archive_id = 1;
data = t2;
}
}
2) No action params in the trigger:
lhsm_archive_trigger {
trigger_on = periodic;
check_interval = 1min;
}
3) The rules section has currently no parameters either ( I had the
archive ids on that level earlier):
lhsm_archive_rules {
ignore_fileclass = empty_files;
ignore_fileclass = no_archive;
rule archive_siteA {
target_fileclass = siteA;
condition { last_mod >= 1min }
# action_params { archive_id = 1;
# data = "t1";
# }
}
rule archive_siteB {
target_fileclass = siteB;
condition { last_mod >= 1min }
# action_params { archive_id = 2; }
}
rule archive_siteAB {
target_fileclass = siteAB;
condition { last_mod >= 1min }
# action_params { archive_id = 3; }
}
# fallback rule
rule default {
action_params { archive_id = 1; }
condition { last_mod >= 1min }
}
}
4) The file class definitions have now the different parameters:
FileClass empty_files {
definition { type == file and size == 0 }
lhsm_archive_action_params { data = "t1"; }
# report = yes (default)
}
FileClass small_files {
definition { type == file and size > 0 and size <= 16MB }
lhsm_archive_action_params { data = "t1"; }
# report = yes (default)
}
FileClass std_files {
definition { type == file and size > 16MB and size <= 1GB }
lhsm_archive_action_params { data = "t2"; }
}
FileClass big_files {
definition { type == file and size > 1GB }
lhsm_archive_action_params { data = "t2"; }
}
FileClass largedir {
definition { type == directory and dircount > 10000 }
}
FileClass siteA {
definition { tree == "/gdatat1/siteA" }
lhsm_archive_action_params { archive_id = 1; }
}
FileClass siteB {
definition { tree == "/gdatat1/siteB" }
lhsm_archive_action_params { archive_id = 2;
data = "t1"; }
}
FileClass siteAB {
definition { tree == "/gdatat1/siteAB" }
lhsm_archive_action_params { archive_id = 3; }
}
FileClass no_archive {
definition { tree == "/gdatat1/no_archive" }
}
The observation is that FileClass definitions small_files, std_files and
big_files does not seem to set the {data} parameter.
This particular configurations does test that the {data} parameter in
principal gets set in a FileClass. If I write a file into /gdatat1/siteB
the file gets migrated to archive_id 2 and data "t1". So both parameters
are overwritten from the default archive_id 1 and "t2".
But if I am writing a small file into /gdatat1/siteAB the {data}
parameter does not get overwritten from the default by the small_files
FileClass.
I did also test a config where I did not set a default and I saw that
the {data} parameter was not set at all and the lfs hsm_archive produced
an error.
I think this looks like a bug...
I have not found a way to implement that logic any other way, but I
would like to hear if there is a different approach.
Thanks,
Gerald
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
robinhood-support mailing list
robinhood-support@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/robinhood-support
