Additional zram + swap tuning

default/systemd/zram-generator.conf.d/90-omarchy.conf

@@ -1,19 +1,9 @@
-# Compressed swap in RAM, sized the way Fedora has shipped it since F34: all of
+# Compressed swap in RAM. zstd averages around 3:1, so even a full device
-# memory, capped at 8G. Fedora weighed the worry that incompressible pages would
+# occupies roughly a third of RAM. Any smaller cap spills reclaim to the disk
-# tie up RAM against five years of desktops reporting no trouble from it, which
+# swapfile, which on encrypted installs pages back through LUKS.
-# is more field evidence than we could gather on our own.
-#
-# The cap is what bounds the downside. A browser-heavy desktop can drive zstd
-# down near 1.1:1, where the device backs far less than it advertises, and 8G
-# keeps that bounded on any machine large enough to notice. The disk swapfile
-# sits behind it at pri=0 for everything past that.
-#
-# Fedora spells this zram-fraction = 1.0 with max-zram-size = 8192; both keys
-# are deprecated in favour of zram-size, which computes the same numbers.
 [zram0]
-zram-size = min(ram, 8192)
+zram-size = ram
 compression-algorithm = zstd
 
-# Above the pri=0 that omarchy-hibernation-setup gives the disk swapfile, so
+# Above the pri=0 that omarchy-hibernation-setup gives the disk swapfile.
-# zram absorbs everything until it's full.
 swap-priority = 100

etc/sysctl.d/99-omarchy-sysctl.conf

@@ -7,8 +7,13 @@ net.ipv4.tcp_mtu_probing=1
 # Anything above 100 tells the kernel that evicting an anonymous page is
 # cheaper than dropping a page-cache page it would have to re-read from disk.
 # With a compressed RAM device that is true, so the disk-era default of 60
-# leaves the page cache starved.
+# leaves the page cache starved. Stop at 150: every page swapped still costs
-vm.swappiness=180
+# a compression now and a decompression on fault-back.
+vm.swappiness=150
+
+# Halve the eagerness to drop dentry/inode cache, which is costly to rebuild
+# and can't spill to zram. Not lower: 0 can OOM.
+vm.vfs_cache_pressure=50
 
 # Read one page per swap-in fault. The default of 8 pays for a seek that zram
 # doesn't have, and every extra page costs a separate decompression.
@@ -21,3 +26,11 @@ vm.watermark_boost_factor=0
 # Keep ~1.25% of memory free instead of 0.1%, so kswapd reclaims in the
 # background rather than letting allocations stall in direct reclaim.
 vm.watermark_scale_factor=125
+
+# Cap dirty pages at 64M/256M instead of the default 10%/20% of RAM, which
+# lets gigabytes of writeback pile up and flush in stalling bursts.
+vm.dirty_background_bytes=67108864
+vm.dirty_bytes=268435456
+
+# With bursts bounded above, the flusher can wake every 15s instead of 5s.
+vm.dirty_writeback_centisecs=1500

etc/tmpfiles.d/omarchy-zswap.conf

@@ -0,0 +1,4 @@
+# Arch's kernel enables zswap by default, which in front of swap-on-zram just
+# double-compresses pages and breaks zramctl accounting. Boot-only (w!), so
+# flipping it on by hand for an experiment sticks until reboot.
+w! /sys/module/zswap/parameters/enabled - - - - N