<div dir="ltr">Hi All,<br>We have set up a new gluster volume, server version 9.3 with clients running 9.0. In this setup we are facing issues where the clients get randomly disconnected and there is no relevant log around that time in client, server and brick logs. One difference we have noticed in this setup vs the other setups which we have in production is that there are many small read/writes happening in this.<br>Wondering if anyone can help with the performance tuning parameters which we can set/unset to optimize our setup.<br>Any suggestions are welcome and appreciated. <br><br>Adding the current server parameters:<br><blockquote style="margin:0 0 0 40px;border:none;padding:0px"><b>Option Value<br></b><b>------ -----<br></b><b>cluster.lookup-unhashed on (DEFAULT)<br></b><b>cluster.lookup-optimize on (DEFAULT)<br></b><b>cluster.min-free-disk 10% (DEFAULT)<br></b><b>cluster.min-free-inodes 5% (DEFAULT)<br></b><b>cluster.rebalance-stats off (DEFAULT)<br></b><b>cluster.subvols-per-directory (null) (DEFAULT)<br></b><b>cluster.readdir-optimize off (DEFAULT)<br></b><b>cluster.rsync-hash-regex (null) (DEFAULT)<br></b><b>cluster.extra-hash-regex (null) (DEFAULT)<br></b><b>cluster.dht-xattr-name trusted.glusterfs.dht (DEFAULT)<br></b><b>cluster.randomize-hash-range-by-gfid off (DEFAULT)<br></b><b>cluster.rebal-throttle normal (DEFAULT)<br></b><b>cluster.lock-migration off<br></b><b>cluster.force-migration off<br></b><b>cluster.local-volume-name (null) (DEFAULT)<br></b><b>cluster.weighted-rebalance on (DEFAULT)<br></b><b>cluster.switch-pattern (null) (DEFAULT)<br></b><b>cluster.entry-change-log on (DEFAULT)<br></b><b>cluster.read-subvolume (null) (DEFAULT)<br></b><b>cluster.read-subvolume-index -1 (DEFAULT)<br></b><b>cluster.read-hash-mode 1 (DEFAULT)<br></b><b>cluster.background-self-heal-count 8 (DEFAULT)<br></b><b>cluster.metadata-self-heal off (DEFAULT)<br></b><b>cluster.data-self-heal off (DEFAULT)<br></b><b>cluster.entry-self-heal off (DEFAULT)<br></b><b>cluster.self-heal-daemon on (DEFAULT)<br></b><b>cluster.heal-timeout 600 (DEFAULT)<br></b><b>cluster.self-heal-window-size 8 (DEFAULT)<br></b><b>cluster.data-change-log on (DEFAULT)<br></b><b>cluster.metadata-change-log on (DEFAULT)<br></b><b>cluster.data-self-heal-algorithm (null) (DEFAULT)<br></b><b>cluster.eager-lock on (DEFAULT)<br></b><b>disperse.eager-lock on (DEFAULT)<br></b><b>disperse.other-eager-lock on (DEFAULT)<br></b><b>disperse.eager-lock-timeout 1 (DEFAULT)<br></b><b>disperse.other-eager-lock-timeout 1 (DEFAULT)<br></b><b>cluster.quorum-type none (DEFAULT)<br></b><b>cluster.quorum-count (null) (DEFAULT)<br></b><b>cluster.choose-local true (DEFAULT)<br></b><b>cluster.self-heal-readdir-size 1KB (DEFAULT)<br></b><b>cluster.post-op-delay-secs 1 (DEFAULT)<br></b><b>cluster.ensure-durability on (DEFAULT)<br></b><b>cluster.consistent-metadata no (DEFAULT)<br></b><b>cluster.heal-wait-queue-length 128 (DEFAULT)<br></b><b>cluster.favorite-child-policy none (DEFAULT)<br></b><b>cluster.full-lock yes (DEFAULT)<br></b><b>cluster.optimistic-change-log on (DEFAULT)<br></b><b>diagnostics.latency-measurement off<br></b><b>diagnostics.dump-fd-stats off (DEFAULT)<br></b><b>diagnostics.count-fop-hits off<br></b><b>diagnostics.brick-log-level INFO<br></b><b>diagnostics.client-log-level INFO<br></b><b>diagnostics.brick-sys-log-level CRITICAL (DEFAULT)<br></b><b>diagnostics.client-sys-log-level CRITICAL (DEFAULT)<br></b><b>diagnostics.brick-logger (null) (DEFAULT)<br></b><b>diagnostics.client-logger (null) (DEFAULT)<br></b><b>diagnostics.brick-log-format (null) (DEFAULT)<br></b><b>diagnostics.client-log-format (null) (DEFAULT)<br></b><b>diagnostics.brick-log-buf-size 5 (DEFAULT)<br></b><b>diagnostics.client-log-buf-size 5 (DEFAULT)<br></b><b>diagnostics.brick-log-flush-timeout 120 (DEFAULT)<br></b><b>diagnostics.client-log-flush-timeout 120 (DEFAULT)<br></b><b>diagnostics.stats-dump-interval 0 (DEFAULT)<br></b><b>diagnostics.fop-sample-interval 0 (DEFAULT)<br></b><b>diagnostics.stats-dump-format json (DEFAULT)<br></b><b>diagnostics.fop-sample-buf-size 65535 (DEFAULT)<br></b><b>diagnostics.stats-dnscache-ttl-sec 86400 (DEFAULT)<br></b><b>performance.cache-max-file-size 0 (DEFAULT)<br></b><b>performance.cache-min-file-size 0 (DEFAULT)<br></b><b>performance.cache-refresh-timeout 1 (DEFAULT)<br></b><b>performance.cache-priority (DEFAULT)<br></b><b>performance.io-cache-size 32MB (DEFAULT)<br></b><b>performance.cache-size 256MB<br></b><b>performance.io-thread-count 16 (DEFAULT)<br></b><b>performance.high-prio-threads 16 (DEFAULT)<br></b><b>performance.normal-prio-threads 16 (DEFAULT)<br></b><b>performance.low-prio-threads 16 (DEFAULT)<br></b><b>performance.least-prio-threads 1 (DEFAULT)<br></b><b>performance.enable-least-priority on (DEFAULT)<br></b><b>performance.iot-watchdog-secs (null) (DEFAULT)<br></b><b>performance.iot-cleanup-disconnected-reqs off (DEFAULT)<br></b><b>performance.iot-pass-through false (DEFAULT)<br></b><b>performance.io-cache-pass-through false (DEFAULT)<br></b><b>performance.quick-read-cache-size 128MB (DEFAULT)<br></b><b>performance.cache-size 256MB<br></b><b>performance.quick-read-cache-timeout 1 (DEFAULT)<br></b><b>performance.qr-cache-timeout 600<br></b><b>performance.quick-read-cache-invalidation false (DEFAULT)<br></b><b>performance.ctime-invalidation false (DEFAULT)<br></b><b>performance.flush-behind on (DEFAULT)<br></b><b>performance.nfs.flush-behind on (DEFAULT)<br></b><b>performance.write-behind-window-size 1MB (DEFAULT)<br></b><b>performance.resync-failed-syncs-after-fsync off (DEFAULT)<br></b><b>performance.nfs.write-behind-window-size 1MB (DEFAULT)<br></b><b>performance.strict-o-direct off (DEFAULT)<br></b><b>performance.nfs.strict-o-direct off (DEFAULT)<br></b><b>performance.strict-write-ordering off (DEFAULT)<br></b><b>performance.nfs.strict-write-ordering off (DEFAULT)<br></b><b>performance.write-behind-trickling-writes on (DEFAULT)<br></b><b>performance.aggregate-size 128KB (DEFAULT)<br></b><b>performance.nfs.write-behind-trickling-writes on (DEFAULT)<br></b><b>performance.lazy-open yes (DEFAULT)<br></b><b>performance.read-after-open yes (DEFAULT)<br></b><b>performance.open-behind-pass-through false (DEFAULT)<br></b><b>performance.read-ahead-page-count 4 (DEFAULT)<br></b><b>performance.read-ahead-pass-through false (DEFAULT)<br></b><b>performance.readdir-ahead-pass-through false (DEFAULT)<br></b><b>performance.md-cache-pass-through false (DEFAULT)<br></b><b>performance.write-behind-pass-through false (DEFAULT)<br></b><b>performance.md-cache-timeout 1 (DEFAULT)<br></b><b>performance.cache-swift-metadata false (DEFAULT)<br></b><b>performance.cache-samba-metadata false (DEFAULT)<br></b><b>performance.cache-capability-xattrs true (DEFAULT)<br></b><b>performance.cache-ima-xattrs true (DEFAULT)<br></b><b>performance.md-cache-statfs off (DEFAULT)<br></b><b>performance.xattr-cache-list (DEFAULT)<br></b><b>performance.nl-cache-pass-through false (DEFAULT)<br></b><b>network.frame-timeout 1800 (DEFAULT)<br></b><b>network.ping-timeout 42 (DEFAULT)<br></b><b>network.tcp-window-size (null) (DEFAULT)<br></b><b>client.ssl off<br></b><b>network.remote-dio disable (DEFAULT)<br></b><b>client.event-threads 12<br></b><b>client.tcp-user-timeout 0<br></b><b>client.keepalive-time 20000<br></b><b>client.keepalive-interval 2000<br></b><b>client.keepalive-count 9<br></b><b>client.strict-locks off<br></b><b>network.tcp-window-size (null) (DEFAULT)<br></b><b>network.inode-lru-limit 50000<br></b><b>auth.allow *<br></b><b>auth.reject (null) (DEFAULT)<br></b><b>transport.keepalive 1<br></b><b>server.allow-insecure on (DEFAULT)<br></b><b>server.root-squash off (DEFAULT)<br></b><b>server.all-squash off (DEFAULT)<br></b><b>server.anonuid 65534 (DEFAULT)<br></b><b>server.anongid 65534 (DEFAULT)<br></b><b>server.statedump-path /var/run/gluster (DEFAULT)<br></b><b>server.outstanding-rpc-limit 128<br></b><b>server.ssl off<br></b><b>auth.ssl-allow *<br></b><b>server.manage-gids off (DEFAULT)<br></b><b>server.dynamic-auth on (DEFAULT)<br></b><b>client.send-gids on (DEFAULT)<br></b><b>server.gid-timeout 300 (DEFAULT)<br></b><b>server.own-thread (null) (DEFAULT)<br></b><b>server.event-threads 12<br></b><b>server.tcp-user-timeout 42 (DEFAULT)<br></b><b>server.keepalive-time 20<br></b><b>server.keepalive-interval 2<br></b><b>server.keepalive-count 9<br></b><b>transport.listen-backlog 1024<br></b><b>ssl.own-cert (null) (DEFAULT)<br></b><b>ssl.private-key (null) (DEFAULT)<br></b><b>ssl.ca-list (null) (DEFAULT)<br></b><b>ssl.crl-path (null) (DEFAULT)<br></b><b>ssl.certificate-depth (null) (DEFAULT)<br></b><b>ssl.cipher-list (null) (DEFAULT)<br></b><b>ssl.dh-param (null) (DEFAULT)<br></b><b>ssl.ec-curve (null) (DEFAULT)<br></b><b>transport.address-family inet<br></b><b>performance.write-behind on<br></b><b>performance.read-ahead off<br></b><b>performance.readdir-ahead off<br></b><b>performance.io-cache off<br></b><b>performance.open-behind on<br></b><b>performance.quick-read on<br></b><b>performance.nl-cache on<br></b><b>performance.stat-prefetch on<br></b><b>performance.client-io-threads on<br></b><b>performance.nfs.write-behind on<br></b><b>performance.nfs.read-ahead off<br></b><b>performance.nfs.io-cache off<br></b><b>performance.nfs.quick-read off<br></b><b>performance.nfs.stat-prefetch off<br></b><b>performance.nfs.io-threads off<br></b><b>performance.force-readdirp true (DEFAULT)<br></b><b>performance.cache-invalidation on<br></b><b>performance.global-cache-invalidation true (DEFAULT)<br></b><b>features.uss off<br></b><b>features.snapshot-directory .snaps<br></b><b>features.show-snapshot-directory off<br></b><b>features.tag-namespaces off<br></b><b>network.compression off<br></b><b>network.compression.window-size -15 (DEFAULT)<br></b><b>network.compression.mem-level 8 (DEFAULT)<br></b><b>network.compression.min-size 0 (DEFAULT)<br></b><b>network.compression.compression-level -1 (DEFAULT)<br></b><b>network.compression.debug false (DEFAULT)<br></b><b>features.default-soft-limit 80% (DEFAULT)<br></b><b>features.soft-timeout 60 (DEFAULT)<br></b><b>features.hard-timeout 5 (DEFAULT)<br></b><b>features.alert-time 86400 (DEFAULT)<br></b><b>features.quota-deem-statfs off<br></b><b>geo-replication.indexing off<br></b><b>geo-replication.indexing off<br></b><b>geo-replication.ignore-pid-check off<br></b><b>geo-replication.ignore-pid-check off<br></b><b>features.quota off<br></b><b>features.inode-quota off<br></b><b>features.bitrot disable<br></b><b>debug.trace off<br></b><b>debug.log-history no (DEFAULT)<br></b><b>debug.log-file no (DEFAULT)<br></b><b>debug.exclude-ops (null) (DEFAULT)<br></b><b>debug.include-ops (null) (DEFAULT)<br></b><b>debug.error-gen off<br></b><b>debug.error-failure (null) (DEFAULT)<br></b><b>debug.error-number (null) (DEFAULT)<br></b><b>debug.random-failure off (DEFAULT)<br></b><b>debug.error-fops (null) (DEFAULT)<br></b><b>nfs.disable on<br></b><b>features.read-only off (DEFAULT)<br></b><b>features.worm off<br></b><b>features.worm-file-level off<br></b><b>features.worm-files-deletable on<br></b><b>features.default-retention-period 120 (DEFAULT)<br></b><b>features.retention-mode relax (DEFAULT)<br></b><b>features.auto-commit-period 180 (DEFAULT)<br></b><b>storage.linux-aio off (DEFAULT)<br></b><b>storage.linux-io_uring off (DEFAULT)<br></b><b>storage.batch-fsync-mode reverse-fsync (DEFAULT)<br></b><b>storage.batch-fsync-delay-usec 0 (DEFAULT)<br></b><b>storage.owner-uid -1 (DEFAULT)<br></b><b>storage.owner-gid -1 (DEFAULT)<br></b><b>storage.node-uuid-pathinfo off (DEFAULT)<br></b><b>storage.health-check-interval 30 (DEFAULT)<br></b><b>storage.build-pgfid off (DEFAULT)<br></b><b>storage.gfid2path on (DEFAULT)<br></b><b>storage.gfid2path-separator : (DEFAULT)<br></b><b>storage.reserve 1 (DEFAULT)<br></b><b>storage.health-check-timeout 20 (DEFAULT)<br></b><b>storage.fips-mode-rchecksum off<br></b><b>storage.force-create-mode 0000 (DEFAULT)<br></b><b>storage.force-directory-mode 0000 (DEFAULT)<br></b><b>storage.create-mask 0777 (DEFAULT)<br></b><b>storage.create-directory-mask 0777 (DEFAULT)<br></b><b>storage.max-hardlinks 100 (DEFAULT)<br></b><b>features.ctime on (DEFAULT)<br></b><b>config.gfproxyd off<br></b><b>cluster.server-quorum-type off
</b><b>cluster.server-quorum-ratio 51<br></b><b>changelog.changelog off (DEFAULT)<br></b><b>changelog.changelog-dir {{ brick.path }}/.glusterfs/changelogs (DEFAULT)<br></b><b>changelog.encoding ascii (DEFAULT)<br></b><b>changelog.rollover-time 15 (DEFAULT)<br></b><b>changelog.fsync-interval 5 (DEFAULT)<br></b><b>changelog.changelog-barrier-timeout 120<br></b><b>changelog.capture-del-path off (DEFAULT)<br></b><b>features.barrier disable<br></b><b>features.barrier-timeout 120<br></b><b>features.trash off (DEFAULT)<br></b><b>features.trash-dir .trashcan (DEFAULT)<br></b><b>features.trash-eliminate-path (null) (DEFAULT)<br></b><b>features.trash-max-filesize 5MB (DEFAULT)<br></b><b>features.trash-internal-op off (DEFAULT)<br></b><b>cluster.enable-shared-storage disable<br></b><b>locks.trace off (DEFAULT)<br></b><b>locks.mandatory-locking off (DEFAULT)<br></b><b>cluster.disperse-self-heal-daemon enable (DEFAULT)<br></b><b>cluster.quorum-reads no (DEFAULT)<br></b><b>client.bind-insecure (null) (DEFAULT)<br></b><b>features.shard off<br></b><b>features.shard-block-size 64MB (DEFAULT)<br></b><b>features.shard-lru-limit 16384 (DEFAULT)<br></b><b>features.shard-deletion-rate 100 (DEFAULT)<br></b><b>features.scrub-throttle lazy<br></b><b>features.scrub-freq biweekly<br></b><b>features.scrub false (DEFAULT)<br></b><b>features.expiry-time 120<br></b><b>features.signer-threads 4<br></b><b>features.cache-invalidation on<br></b><b>features.cache-invalidation-timeout 600<br></b><b>ganesha.enable off<br></b><b>features.leases off<br></b><b>features.lease-lock-recall-timeout 60 (DEFAULT)<br></b><b>disperse.background-heals 8 (DEFAULT)<br></b><b>disperse.heal-wait-qlength 128 (DEFAULT)<br></b><b>cluster.heal-timeout 600 (DEFAULT)<br></b><b>dht.force-readdirp on (DEFAULT)<br></b><b>disperse.read-policy gfid-hash (DEFAULT)<br></b><b>cluster.shd-max-threads 1 (DEFAULT)<br></b><b>cluster.shd-wait-qlength 1024 (DEFAULT)<br></b><b>cluster.locking-scheme full (DEFAULT)<br></b><b>cluster.granular-entry-heal no (DEFAULT)<br></b><b>features.locks-revocation-secs 0 (DEFAULT)<br></b><b>features.locks-revocation-clear-all false (DEFAULT)<br></b><b>features.locks-revocation-max-blocked 0 (DEFAULT)<br></b><b>features.locks-monkey-unlocking false (DEFAULT)<br></b><b>features.locks-notify-contention yes (DEFAULT)<br></b><b>features.locks-notify-contention-delay 5 (DEFAULT)<br></b><b>disperse.shd-max-threads 1 (DEFAULT)<br></b><b>disperse.shd-wait-qlength 1024 (DEFAULT)<br></b><b>disperse.cpu-extensions auto (DEFAULT)<br></b><b>disperse.self-heal-window-size 32 (DEFAULT)<br></b><b>cluster.use-compound-fops off<br></b><b>performance.parallel-readdir on<br></b><b>performance.rda-request-size 131072<br></b><b>performance.rda-low-wmark 4096 (DEFAULT)<br></b><b>performance.rda-high-wmark 128KB (DEFAULT)<br></b><b>performance.rda-cache-limit 10MB<br></b><b>performance.nl-cache-positive-entry false (DEFAULT)<br></b><b>performance.nl-cache-limit 10MB<br></b><b>performance.nl-cache-timeout 600<br></b><b>cluster.brick-multiplex disable<br></b><b>glusterd.vol_count_per_thread 100<br></b><b>cluster.max-bricks-per-process 250<br></b><b>disperse.optimistic-change-log on (DEFAULT)<br></b><b>disperse.stripe-cache 4 (DEFAULT)<br></b><b>cluster.halo-enabled False (DEFAULT)<br></b><b>cluster.halo-shd-max-latency 99999 (DEFAULT)<br></b><b>cluster.halo-nfsd-max-latency 5 (DEFAULT)<br></b><b>cluster.halo-max-latency 5 (DEFAULT)<br></b><b>cluster.halo-max-replicas 99999 (DEFAULT)<br></b><b>cluster.halo-min-replicas 2 (DEFAULT)<br></b><b>features.selinux on<br></b><b>cluster.daemon-log-level INFO<br></b><b>debug.delay-gen off<br></b><b>delay-gen.delay-percentage 10% (DEFAULT)<br></b><b>delay-gen.delay-duration 100000 (DEFAULT)<br></b><b>delay-gen.enable (DEFAULT)<br></b><b>disperse.parallel-writes on (DEFAULT)<br></b><b>disperse.quorum-count 0 (DEFAULT)<br></b><b>features.sdfs off<br></b><b>features.cloudsync off<br></b><b>features.ctime on<br></b><b>ctime.noatime on<br></b><b>features.cloudsync-storetype (null) (DEFAULT)<br></b><b>features.enforce-mandatory-lock off<br></b><b>config.global-threading off<br></b><b>config.client-threads 16<br></b><b>config.brick-threads 16<br></b><b>features.cloudsync-remote-read off<br></b><b>features.cloudsync-store-id (null) (DEFAULT)<br></b><b>features.cloudsync-product-id (null) (DEFAULT)<br></b><b>features.acl enable<br></b><b>cluster.use-anonymous-inode yes</b></blockquote><br clear="all"><div><br></div>-- <br><div dir="ltr" class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr"><div dir="ltr"><div dir="ltr"><font face="arial, helvetica, sans-serif">Regards,<br>Shreyansh Shah</font><br><div style="margin:0in 0in 0.0001pt"><font><font color="#0b5394" style="font-weight:bold">Alpha</font><font color="#666666" style="font-weight:bold">Grep</font><font color="#000000"><b> Securities Pvt. Ltd.</b><br></font></font></div></div></div></div></div></div>