[sssd PR#5937][opened] [WIP] Another attempt to resolve #5134

[alexey-tikhonov]

   URL: https://github.com/SSSD/sssd/pull/5937
Author: alexey-tikhonov
 Title: #5937: [WIP] Another attempt to resolve #5134
Action: opened

PR body:
"""
There are two ideas:
 - get rid of re-alloc for every new member added: instead reserve space with a margin and
shrink later.
 - don't add every new member to the list immediately: instead keep new members in a
temporarily hash table. This will allow to avoid cross strcmp() at least among new members
- hopefully, hash table should be much faster.


2nd patch actually makes 1st patch obsolete, but I will keep it a while to measure
impact.



This is "WIP" as I didn't conduct perf measurements yet.

"""

To pull the PR as Git branch:
git remote add ghsssd https://github.com/SSSD/sssd
git fetch ghsssd pull/5937/head:pr5937
git checkout pr5937

[sssd PR#5937][+Bugzilla] [WIP] Another attempt to resolve #5134

[alexey-tikhonov]

  URL: https://github.com/SSSD/sssd/pull/5937
Title: #5937: [WIP] Another attempt to resolve #5134

Label: +Bugzilla

[sssd PR#5937][comment] [WIP] Another attempt to resolve #5134

[alexey-tikhonov]

  URL: https://github.com/SSSD/sssd/pull/5937
Title: #5937: [WIP] Another attempt to resolve #5134

alexey-tikhonov commented:
"""
RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=1893192
"""

See the full comment at https://github.com/SSSD/sssd/pull/5937#issuecomment-1002782096

[sssd PR#5937][comment] [WIP] Another attempt to resolve #5134

[alexey-tikhonov]

  URL: https://github.com/SSSD/sssd/pull/5937
Title: #5937: [WIP] Another attempt to resolve #5134

alexey-tikhonov commented:
"""
I did actual testing and conclusions are:
 - I correctly identified the bottleneck:
https://github.com/SSSD/sssd/blob/4897c28741112b547a69feb7c887764c64cc954...
 - my patch(-es) doesn't resolve the issue because by the moment `Received 49999
dereference results` `members->num_values` is already 49999 (i.e. all those members are
already known)
"""

See the full comment at https://github.com/SSSD/sssd/pull/5937#issuecomment-1006932032

[sssd PR#5937][comment] [WIP] Another attempt to resolve #5134

[alexey-tikhonov]

  URL: https://github.com/SSSD/sssd/pull/5937
Title: #5937: [WIP] Another attempt to resolve #5134

alexey-tikhonov commented:
"""
I did actual testing and conclusions are:
 - I correctly identified the bottleneck:
https://github.com/SSSD/sssd/blob/4897c28741112b547a69feb7c887764c64cc954...
 - my patch(-es) doesn't resolve the issue because (in certain configs) by the moment
`Received 49999 dereference results` `members->num_values` is already 49999 (i.e. all
those members are already known)
"""

See the full comment at https://github.com/SSSD/sssd/pull/5937#issuecomment-1006932032

[sssd PR#5937][comment] [WIP] Another attempt to resolve #5134

[alexey-tikhonov]

  URL: https://github.com/SSSD/sssd/pull/5937
Title: #5937: [WIP] Another attempt to resolve #5134

alexey-tikhonov commented:
"""
I did actual testing and conclusions are:
 - I correctly identified the bottleneck:
https://github.com/SSSD/sssd/blob/4897c28741112b547a69feb7c887764c64cc954...
 - my patch(-es) doesn't resolve the issue because (in certain (or most?) configs) by
the moment `Received 49999 dereference results` `members->num_values` is already 49999
(i.e. all those members are already known)
"""

See the full comment at https://github.com/SSSD/sssd/pull/5937#issuecomment-1006932032

[sssd PR#5937][comment] [WIP] Another attempt to resolve #5134

[alexey-tikhonov]

  URL: https://github.com/SSSD/sssd/pull/5937
Title: #5937: [WIP] Another attempt to resolve #5134

alexey-tikhonov commented:
"""
Pavel mentioned that possible reason for the check might be
https://docs.microsoft.com/en-us/previous-versions/windows/desktop/ldap/s...
"""

See the full comment at https://github.com/SSSD/sssd/pull/5937#issuecomment-1012077810