Newwebs Techs logo Newwebs TechsBuilding websites, explained for beginners
Hosting and Domains

Inode and Entry Process Limits on Shared Hosting

A reader wrote to me last March after his host suspended his site at nine on a Friday evening. The control panel said "resource limit reached" and nothing else.

A hosting control panel resource usage page showing a file usage bar near its ceiling beside a seven day entry process graph

A reader wrote to me last March after his host suspended his site at nine on a Friday evening. The control panel said "resource limit reached" and nothing else. He checked his transfer figure: 11 GB used on a plan that advertised unlimited bandwidth. He checked disk: 4.2 GB out of 20. Nothing looked close to full, and the site kept returning a 508 error for the better part of two hours until support answered.

The limits that knock small sites offline are almost never the ones printed on the sales page. They are inodes, entry processes and disk input output, and most shared plans enforce all three while mentioning none of them until you trip one. In his case a backup plugin had been writing a nightly archive since early 2024 and had left roughly 340,000 small files in one folder. Deleting that folder brought the site back in four minutes. Nobody had told him the files counted for anything.

What an inode actually counts

An inode is one entry in the filesystem index. Every file counts as one. Every folder counts as one. Every email message sitting in a mailbox on that account counts as one, and so does every cached thumbnail, every session file, every log line that got its own file. Size does not enter into it. A 4 GB video is one inode. A WordPress install with 2,000 plugin files and 9,000 generated image sizes is 11,000.

Typical shared plans sit between 100,000 and 400,000 inodes. That sounds enormous until you look at where they go. A plain WordPress site with a dozen plugins starts around 25,000 files. Enable six image sizes and upload 500 photos and you have added 3,000 more. Put an uncompressed daily backup in a folder under the web root and you are adding tens of thousands a week.

Entry processes are a different thing and get confused with it constantly. An entry process is one request your server is executing right now. Shared plans commonly allow 20 concurrent, sometimes 10 on the cheapest tier. That number is not about traffic over a day. It is about how many requests pile up at the same instant, which means a slow site hits the ceiling long before a fast one with the same visitor count.

The numbers worth writing down

Here is the shape of what a cheap plan usually gives you, drawn from the four hosts I have accounts on. Figures move around, so read yours rather than trusting mine.

LimitCommon allowanceWhat eats it firstSymptom when you hit it
Inodes150,000 to 300,000Backup folders, old mail, image sizesUploads fail, cron stops, account read only
Entry processes10 to 20 concurrentSlow queries, bots, uncached pages508 error for some visitors, not all
I/O throughput1 to 10 MB per secondUnindexed database tables, log writesSite feels fine then crawls for a minute
Physical memory768 MB to 2 GBPage builders, image resizingWhite screen on one page only
CPU100 percent of one coreSearch pages, feed readersResponse time jumps from 400 ms to 9 s

The pattern worth noticing is that four of those five produce intermittent symptoms. That is why they are so hard to diagnose from the outside. You load the page, it works, you conclude the problem fixed itself, and three hours later someone else gets the error. If you have ever had a visitor insist your site was down on a day your monitoring said it was up, this family of limits is the usual explanation. Uptime monitoring on a one minute interval will miss most of these entirely, because the error window is often shorter than the gap between checks.

Why upgrading is usually the wrong first move

The standard advice when you see a resource error is to move up a tier, and hosts are happy to agree with it. I think that is wrong most of the time, and I say this having paid for an upgrade I did not need in 2023.

An inode ceiling is a housekeeping problem, not a capacity problem. Doubling your allowance to 300,000 when a plugin is adding 40,000 files a week buys you about two months. The underlying growth does not stop, and you have now added a monthly cost to something a single deletion would have solved. The same goes for entry processes. If one uncached page takes 3.5 seconds to render, each visitor holds a slot for 3.5 seconds, and ten visitors arriving in the same four second window will exhaust a 10 process limit. Caching that page down to 80 ms means the same ten visitors hold slots for less than a second in total.

There are real reasons to leave shared hosting, and I covered the honest ones in shared hosting versus a VPS. Running out of inodes is not on that list. Sustained CPU use from genuine traffic is. A database that has grown past what a shared MySQL instance will keep in memory is. A plugin filling a folder with garbage is a Tuesday afternoon job, not a reason to pay more every month.

The exception I will grant: if your host counts mailboxes against the same inode pool as your website, and you genuinely need 40 mailboxes, splitting mail onto a separate provider is worth the money. Mixing those two on one quota causes a particular kind of misery where a full inbox stops your contact form from writing to the database.

Finding your own numbers in about ten minutes

Most panels bury this. In cPanel it sits under Statistics in the sidebar, sometimes labelled File Usage rather than inodes, and in CloudLinux panels there is a Resource Usage page with a seven day graph that is far more useful than the current figure.

Open the resource usage page and note the inode count, then come back in a week and note it again. The slope matters more than the number. Growth of 500 a week is normal for a site that gets edited. Growth of 20,000 a week means something automated is misbehaving, and the graph will tell you that faster than any support ticket.

Then find the heavy folders. Over SSH, du -h -d 1 ~ lists the top level by size, and repeating it one level down narrows the search quickly. In a file manager, sort by item count rather than size if your panel offers it. Check the four usual offenders before anything else: a plugin backup directory, an error log that has been growing since launch, a cache folder nobody clears, and mail in accounts you forgot you created.

For entry processes, read the graph for your busiest hour rather than the daily average. If the line touches your ceiling even once a day, some visitors are already seeing errors you have not heard about. Write both figures and the date into the same note where you keep hosting details. In six months you will want the comparison and you will not remember.

On the deletion itself, do the obvious safety step first. Download a copy of whatever you are about to remove, or at least confirm a working restore exists, which is a habit I argue for in backing up a site and testing the restore. I have watched someone delete a folder named backups that was the only copy of a client site, and the folder name was not a lie, it was just the last copy.

One more thing about logs. An error log that grows by 200 MB a month is telling you something. Read the last hundred lines before you delete it. Half the time the repeated warning in there is also the reason your pages are slow, which means one afternoon fixes both the inode problem and the entry process problem. That is the closest thing to a free win in this area.

FAQ

Does deleting files free inodes immediately?

Usually within a few minutes, though some panels only recalculate on a schedule and can show the old figure for up to 24 hours. If the count has not moved after a day, check whether the files went to a trash folder on the same account, which happens with file managers and with mail clients set to keep deleted messages on the server.

Do image thumbnails really matter that much?

Yes, and more than people expect. A theme that registers eight image sizes turns one upload into nine files. Five hundred photos become 4,500 inodes. Trimming unused sizes and regenerating is one of the few actions that frees tens of thousands of entries without deleting anything a visitor can see.

How many concurrent visitors can 20 entry processes handle?

There is no honest single answer, because it depends entirely on how long each request takes. As a rough feel, a cached page served in 100 ms lets one process handle around ten requests per second, so 20 processes cope with a few hundred simultaneous readers. The same 20 processes on a 3 second page fall over at around seven.

Will a CDN help with these limits?

It helps with entry processes and I O, since requests for images and scripts never reach your server. It does nothing for inodes, because the files still live on your disk. Caching full HTML pages at the edge helps far more than caching assets alone, and the hosting section has more on where that line sits.

FN
Felix Nkemelu

Felix covers website speed and hosting for beginners, turning technical topics into practical decisions.

More posts by Felix

More in Hosting and Domains

01doorganico 084goad 3hot8 4465456 after5internprogram ahandfulofblessings ahmadpc ahomcybersecurity allodialresources annodex antiquesandperiodcollectibles aotc asianworldkashmir beedlectrics bloggertemplatefree bluefeatheragency bodyweightsolution campaign08blog carfixadvisor catstreetcollective cbstvsports cnfanspurchasing committeeof500yearsofdignityandresistance covertopsdjs cs-kaiyuangame cuestacomunicaciontotal cxoutlet dancebetgame deskgearhub destinyvillastay detectmagazine di33 diamondsteelbuildings donmo dyanadropship emarsysfinance emimattl escapechan eskitarih evautoplay everythingweddings-orlando farmzone feedviewerapp foxstudiosscratch gamevui123 gleasongrade greetingsday hacklabasu haskell iw-kaiyun jagoanproperti jilidskripsi joltbusinessfinance juegodepokergratis juegospucca kaeru-festival latestmehndidesigns lestanzediottaviano litemagazines localcommunityhero lustrsazan mariashebets menswheels nikkit ootheday oxigenvpn plurfitness prefetial probudgetcarpetcleaning psychicmrsmartin pyaten qvantor rejoignezlogis rexzbot riselayouts ruchin rwxqsgd5 sakaltime shibiaowang shop-bot skrtips spotlesshaven strcoding tetapatrium topselller uneplanteartificielle vipertechcr webtech1223411 wo-kygaming yamaha-mekar yeknava yoybuysheets zenunwa yamaha-mekar jilidskripsi ankertoto ankertoto ankertoto